A non-computer approach to the critical path method for the construction

industry.
Fondahl, John W
Stanford, Calif., Dept. of Civil Engineering, Stanford University, 1962.

http://hdl.handle.net/2027/mdp.39015000453970
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

Public Domain, Google-digitized

http://www.hathitrust.org/access_use#pd-google
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

This work is in the Public Domain, meaning that it is

not subject to copyright. Users are free to copy, use,
and redistribute the work in part or in whole. It is possible
that heirs or the estate of the authors of individual portions
of the work, such as illustrations, assert copyrights over
these portions. Depending on the nature of subsequent
use that is made, additional rights may need to be obtained
independently of anything we can address. The digital
images and OCR of this work were produced by Google,
Inc. (indicated by a watermark on each page in the
PageTurner). Google requests that the images and OCR
not be re-hosted, redistributed or used commercially.
The images are provided for educational, scholarly,
non-commercial purposes.
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 —a. .
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

?ROPERTT OV
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

ARTES SC1ENTIA VERITAS

 I-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
November 1961

Technical Report No. 9

Revised 1962

A NON-COMPUTER APPROACH

to the

CRITICAL PATH METHOD

for the

CONSTRUCTION INDUSTRY

2nd Edition

JOHN W. FONDAHL

Associate Professor of Civi1 Engineering

PREPARED UNDER

RESEARCH CONTRACT NBy-17798

BUREAU OF YARDS AND DOCKS, U.S. NAVY

Distributed by

THE CONSTRUCTION INSTITUTE

Department of CIVIL ENGINEERING

STANFORD UNIVERSITY
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
A NON-COMPUTER APPROACH

to the

CRITICAL PATH METHOD

for the

CONSTRUCTION INDUSTRY

2nd Edition

John U. Fondahl

Associate Professor^of Civil Engineering

Prepared Under Contract No. NBy-l7798

for the Bureau of Yards and Docks,

U. S. Navy

Department of Civil Engineering

Stanford University

Stanford, California

l962

303
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Department of Civil Engineering

Stanford University

Stanford, California

© l96l, l962 by John W. Fondahl

All rights reserved

Printed in the United States of America

Reproduction in whole or in part is permitted

for any purpose of the United States Government.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
INDEX

Summary 3

Introduction 5

Historical Development l3

The CPM Approach - General l5

Phase I l8

Phase II 25

Phase III - General 3l

Phase III - Normal Start 39

Phase III - All-Crash Start 58

Phase III - Conventional Estimate Start - Mechanics 63

Phase III - Conventional Estimate Start - Advantage 67

Evaluation and Application 70

Future Implications 76

Bibliography 79

Acknowledgement 85

Appendixes of Charts and Tables

Phase I - Project Network Appendix A

Phase II - Construction Schedule Appendix B

Phase III - Scheduling Variations - General Appendix C

Phase III - "Normal Start" Procedure Appendix D

Phase III - "All-Crash Start" Procedure Appendix E

Phase III - "Conventional-Estimate Start

Procedure Appendix F

Optimum Solution Exception Appendix G

An Informal Practical Approach to the

Application of CPM Appendix H

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
SUMMARY

The Critical Path Method for planning, scheduling, and control of

project operations is a new and useful tool now becoming available to

the construction industry. Successful applications have already proven

its value. To date, except for the simplest of cases, the use of this

technique has been largely dependent on programmed solutions by elec-

tronic computers. This report presents the mechanics of noncomputer

methods for applying the Critical Path Method. These employ the same

input data and furnish the same output information a6 the computer

methods. In addition, alternate approaches to the solution of the

problem are presented which permit the important scheduling variation

phase to begin with a normal estimate rather than a 6et of data for

an artificial condition.

There are three goals toward which this report is aimed. The

first is to present a noncomjniter method for obtaining the benefits

of critical path scheduling that is practical to apply to many of

the projects encountered by the construction contractor. The

is to develop the possibilities inherent in a step-by-step, manual

solution to overcome some of the shortcomings of computer programmed

solutions. The third is to offer the reader an opportunity to under-

V/

stand the details of the method and the assumptions upon which it is

based, by discussing them and presenting a complete solution to an

illustrative problem.

The solutions of a complex problem by computer methods, elimin-

ating tedious calculations and possible errors, is a valuable step

forward. This report is not intended to oppose such methods but,

rather, to offer a stepping stone between conventional procedures and

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

these more sophisticated practices. The need for such a stepping

stone is justified on two bases. First, many potential users find

it inconvenient to use electronic computers or are not yet "computer

conscious." Second, the computer approaches are not completely sat-

isfactory in all respects. It is anticipated that a broader acquaint-

ance with, and use of, the Critical Path Method, made possible by

-3-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
noncomputer methods, will lead eventually to an even wider employment of

computer techniques as well as the development of improvements in them.

This report is not prepared for the casual reader. It attempts to

present sufficiently detailed explanations of the procedures in as con-

cise a manner as practical for the benefit of those who will actually

apply these methods. Unless these details are followed closely and the

effort is made to perform the calculations independently, using the ex-

amples given as a guide and check, the report will have limited value.

Some of the procedures that seem complex upon initial examination will

appear much simpler after a few trials are attempted. It should be

remembered that although a project subdivided into a small number of

operations is used to illustrate the mechanics of the methods, the pro-

cedures are designed with the intention that they may be applied to

projects having a considerably larger number of operations.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
INTRODUCTION

Planning, scheduling, and control of activities are functions re-

quired by most industries. A new and improved technique for performing

these functions has been developed. It has been discussed under several

titles including Critical Path Method (CPM),l Critical Path Scheduling

(CPS),^ Critical Path Analysis (CPA),3 and Least Cost Estimating and

Scheduling (LESS). The use of the designation Critical Path Method,

and the abbreviation CPM, has been adopted in this report. It appears

to be the most widely used designation and does not imply a limitation

in the use of the technique to only scheduling or estimating.

This report is concerned with the application of the Critical Path

Method to the planning, scheduling, and control of construction projects

by construction contractors. Most of the procedures discussed are as

applicable to other industries, activities, and users as those chosen

for this presentation. For example, maintenance jobs performed by an

owner's own forces may receive similar analysis. However, the impor-

tance of the construction industry justifies the examination of any new

tool for possible cost reduction from its standpoint alone. Moreover,

a desire to be specific makes the selection of one potential user's

problems a more satisfactory setting for discussion.

A construction project may be subdivided into a number of separate

steps, or operations, necessary for its accomplishment. Each one of

these operations may be performed by many possible combinations of

methods, labor skills, crew sizes, equipment, and working hours. A

major factor in choosing the best combination, is cost and, on first

glance, it would seem desirable to perform each operation at its least

cost in order that the overall project may have the lowest total cost.

'"Perini Corp. Pioneers CPM, New Tool for Job Management." Engineering
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

News Record, January 26, l96l

'Sayer, J.S., Kelley, J.E., Jr., and Walker, M.R. , "Critical Path Sched-

uling," Factory, July, l960

'Clarke, R.W., "An Introduction to Critical Path Analysis." Graduate

School of Business, Stanford University, March, l96l

* Backer, F., Jr., "A Discussion of Problems Involved in LESS (Least

Cost Estimating and Scheduling)," IBM Applied Science, Dallas, Texas,

l960
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

-5-
In a broader sense, however, there are other costs associated with pro-

ject completion. From the owner's standpoint, these inc'.ude the costs

or lost revenue .o be saved by early completion and utilization of the

facility being constructed. Such costs are sometimes anticipated in

advance when liquidated-damage or bonus-penalty clauses are included

in construction contracts. These factors form the basis for establish-

ing completion dates that may make the performance of the project at

least direct cost an impossibility. From the contractor's standpoint,

eve^JjicxejiSing^indirect or overhead costs, anticipated labor and

material cost increases, and the early release of key personnel and

equipment for other work, are factors that make project scheduling for

lowest direct cost unattractive...

It is only by considering all the cost items connected with pro-

ject performance and utilization that the best plan of action can be

determined. This introduces time as an essential variable since almost

alX,casts vary.with time.- But they vary with_ time in opposing directions.

Direct costs tend to decrease as more time is permitted for performance.

On the other hand, the costs associated with the duration of the project

tend to increase as more time is taken for performance. It,follows that

there is some balance between direct cost and time that offers the best

solution. For construction contracts, the owner attempts to find this

balance point when he specifies the completion time, and the contractor

attempts to find it when he plans his detailed schedule of operations.

The recognition of the joint and interrelated importance of both

cost and time in project planning and in project control as well is not

a new concept. For many years appraisals based upon good judgment

assisted by limited trial and error solutions have been employed. How-

ever, systematic methods for considering the joint and interrelated

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

effects of cost and time in reaching a schedule that gives lowest overall

project cost is a new concept.

The problem is not a simple one. Each operation into which a pro-

ject is divided can be performed in different ways at different costs

and in different lengths of time. The problem would be simple if the

overall project time had no importance. In this case, each operation

would be performed in the manner that produced the least direct cost.

-6-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
For reasons already discussed, this would be a rare situation. Overall

project time must generally be reduced from that corresponding to least

direct cost to arrive at a more favorable solution. This is accomplish-

ed by speeding up some of the operations. Speeding up an operation

almost always results in higher direct costs since such measures as over-

time or shift work, oversized crews with less effectiveness, larger

equipment, and different, more expensive methods are generally required.

To produce the same time decrease in one operation as another generally

requires a different additional expenditure. The solution to the problem

would still be relatively simple if the operations formed successive

links in a single chain of events through the project. Then the project

time could be decreased to the desired duration by progressively shorten-

ing those operations which could be altered at least increase in cost.

However, this is also a rare situation. Instead of successive links in

a single chain of events, there is normally a complex network of concurrent,

overlapping and interrelated operations. Shortening a single operation

may only serve to increase project cost without decreasing project time

since other controlling operations have not been shortened. At the other

extreme, shortening all operations simultaneously will increase project

cost more than is necessary to obtain a corresponding decrease in project

time. There are usually certain combinations of operations that may be

shortened to produce the most economic project shortening. In other

instances, shortening a single operation even though this is expensive

may offer a better solution than shortening another operation that alone

is less expensive but which requires concurrent shortening of other

operations. Many such selections must generally be made in arriving

at a project schedule. Without a systematic method on which to base

these decisions, it is unlikely that the unique combination giving the

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

best answer will be obtained.

The Critical Path Method offers this systematic approach. It also

offers additional information for project control and other purposes

beyond that provided by the more conventional project schedules as

commonly represented by bar charts. Specific examples of the information

provided by CPM and some of the resulting applications are as follows:

-7-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
'(l) It pinpoints the operations whose completion times are

responsible for establishing the overall project duration.

With these critical operations clearly identified, major

attention may be directed toward keeping them on schedule

in order that the planned completion date may be met.

2) It gives a quantitative evaluation of the amount of lee-

way, or "float," that each of the other operations possess.

Within the limits of float time, these non-critical

operations may be started later or completed more slowly

than the original schedule indicated, without detrimental

effect on the project completion date. The importance

of variations from the planned timetable of events is

more clearly indicated to management if these limits of

float are known. Moreover, boundaries are established

within which operation times may be shifted to smooth out

manpower or equipment requirements. This is desirable

in the job planning stage.

(3) It shows the most economical scheduling for all operations

for each possible project completion date. This is valuable

planning information that permits correct consideration of

both time and cost in choosing methods, equipment, materials,

crews, and work hours. It replaces the trial and error

juggling of operation times to arrive at a feasible solution

which would seldom be the best possible solution.

(4) It provides the necessary-data for choosing the best project

completion date. The results from (3) may be plotted on a

time-cost graph along with curves for indirect costs and

bonus-penalty or liquidated damage payments. The sum of

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

the curves would produce a total cost curve having a low-

est cost point that would indicate the optimum project

duration.

* The term "float" has been widely adopted as a measurement of operation

scheduling leeway.

-8-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
(5) It offers a means for assc ;ing the rffect on the overalI

project of variations in oni' operation or the addition of

an operation or operations. Change orders and extra work

orders often affect the performance of more than the op-

erations directly involved. Lack of a rational basis for

determining this effect leads to frequent arguments and

sometimes to inequitable settlements. The CPM analysis can

indicate which other operations are affected and what the

overall effect is, if any, on the project completion time.

If the change would extend completion time but the owner

desires that work be accelerated to meet an established

completion date, CPM offers a means for rescheduling the i

operations still to be completed at the least increase in

costs. Furthermore, such time and cost effects can be

computed at any stage of contract completion.

The Critical Path Method has, in its short history, been successfully

applied by a number of users in a variety of industries. These include

both building and heavy construction. Many individuals and groups have

worked, and are currently working, on its development and improvement.

Its mathematical basis is an operations research procedure known as para-

metric linear programming.* The process is a complex one involving many

computational steps. It is not surprising that the methods developed

employ electronic computers. At least three of the major computer man-

ufacturers have standard programs for solving various CPM problems, and

they are working on others.

The work on which this report is based involved an independent

effort to develop the mechanics necessary for applying the Critical

Path Method. It was also aimed at the application of these mechanics

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

by "manual" methods. While there are many disadvantages to employing

a complex method without the aid of an electronic computer, there are

also many justifications for a manual procedure. These justifications

are possibly temporary ones and may disappear when computers become a

'Kellcy, J.E. Jr., "Critical-Path Planning and Scheduling: Math-

ematical Basis," Operations Research, Vol. 9, No. 3, May-June, 1961.

-9-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
more common tool in contractor's offices and better programs are devised

that can more satisfactorily handle contractor's problems. It is quite

possible that the use of CPM based on computer methods will eventually

be almost universally adopted by construction contractors. Because such

a movement is already underway, it might seem a disservice to recommend

a noncomputer method, since the elimination of routine and tedious cal-

culations by computers serves to release more time for the application

of judgment and to decrease the probability of errors. However, in this

instance, introduction of noncomputer methods is not considered a step

backward. Rather, it is a desirable intermediate step between those

methods almost universally in current use and the more sophisticated ones

that are becoming available. The need for such a stepping stone can be

justified by the two conditions, already mentioned, described in more

detail as follows:

(l) Limited current use of computers in the industry:

Although a number of the larger organizations have

employed specialists and established computer centers,

most smaller companies and most field organizations do

.Snot have such facilities. It is usually true that they

could make use of computer service centers and, in many

cases, home office facilities. However, this is either

inconvenient or there is a lack of appreciation of the

benefits that are available. A manual approach might

i , encourage a much more widespread application of CPM for

v,-* problems of moderate complexity involving original pro-

1' ject planning, subsequent revisions to project planning,

project control, and contract change settlements. As

familiarity with the methods grows and more complex

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

problems are undertaken, the use of computers would

naturally become a logical step to take.

(2) Shortcomings of computer programs developed to date in

offering a completely satisfactory approach to the problem:

All methods for solving the problem of schedule vari-

ations require simplifying approximations for the operation

-l0-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
time-cost relationships. But the computer methods require

that these approximations apply over wider ranges. Moreover,

they are much less flexlMe in allowing revisions of the

input data as the schedule takes shape. A step-by-step

manual method allows the planner to retain more judgment

control in making change" ir: the input data. It also per-

mits him to pick a starcmg point that limits the range

over which approximations must apply. With the computer,

all the input data must be supplied at the outset, while

with the manual approach the introduction of new data is

allowed as the schedule deve ops. The computer approach

assumes that each operation is independent except for the

sequential relationships indicated by the project network.

Actually, cost interactions between operations frequently

result when schedule changes are made. The manual approach

permits the planner to experiment with such changes that

may affect several operations rather than just one.

As more experience in applying CPM mechanics to con-

struction problems is developed, improved computer programs

that remedy such shortcomings will probably be developed.

Against these justifications for a manual approach to the problem,

there are some very obvious disadvantages, including:

(l) A considerable amount of computational and data recording

effort, even for problems of little complexity.

(2) The favorable opportunity for error involved in any long,

routine set of calculations.

(3) The much longer time to perform the computational work.

(A) The reluctance to make changes in the project network

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

that would require much of the computational work to be

repeated.

(5) The inconvenience of frequently updating the schedule which

also may require repeat performance of a considerable por-

tion of the computational work.

To some extent these disadvantages may be lessened by such measures as:

-ll-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
(l) Developing methods for checking computed results.

(2) Developing and using flexible, standard printed forms on

which to perform computational steps.

(3) Developing procedures for easy updating of data.

These might include:

(a) Data presentation on peg boards

(b) Data presentation on chalk boards

(c) Data arrangement by card file to permit simple

changes in precedence order

(4) Carrying through only an incomplete but sufficient solution

of the problem. Considerable work is involved in performing

the schedule variation phase of the method. While this pro-

cedure may be worked through many cycles, each cycle repre-

sents an improved schedule, and only a few cycles may give

the optimum schedule. Therefore the entire process usually

need not be carried to completion.

(5) Utilizing a well conceived, conventionally estimated schedule

as a starting point. The computer programs usually start

at the extreme end of the overall project cost-time curve

and proceed to shorten the project from that point.

Starting with a feasible schedule for the contract time

allotted, or considered reasonable, a very few cycles of

calculations may be required to obtain the optimum

schedule.

Before considering the manual method to be presented, a few remarks on

the history of the development of the Critical Path Method are in order.

-l2-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
HISTORICAL DEVELOPMENT

The basic concepts and detailed procedures for the Critical Path

Method have been developed since l956. Two pioneering efforts resulted

in two parallel, but different, approaches. These two can be appropriate

ly designated as the "probahitistic" approach and the "deterministic"

approach. The probabilistic method first appeared as a contract control

system known as PERT (Program Evaluation Research Task). It was devel-

oped under the sponsorship of the Special Projects Office, Bureau of

Ordnance, U.S. Navy, for use as a control tool over contracts for the

Fleet Ballistic Missile Program. The various contracts included in

such a program correspond to the operations making up a single construc-

tion project. These contracts involved, to a large extent, research

and development work and the manufacture of component parts that had

never before been built. Hence neither time nor cost could be esti-

mated accurately. Instead, completion times were treated on a prob-

abilistic basis. Contractors were asked to predict an optimistic, a

most likely, and a pessimistic estimate of time requirements. A

formula was devised for weighing these estimates to determine an ex-

pected time of completion for each contract. These times were then

incorporated into a control system. This system did not consider cost

as a variable. Other more recent work has been reported, known as

PERTCO (PERT with costs), to include cost data on the same sort of

probabilistic basis. The PERT system has been adopted by the Air Force

and its name modified to PEP (Program Evaluation Procedure). It is

being used by the Air Force and the Army in connection with various

missile programs.

The deterministic case is much more suitable for construction ap-

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

plications than is the probabilistic case, since it is used when cost

as well as time is to be considered as a controlling variable. It

""Better Plans Come From Study of Anatomy of an Engineering Job,"

Business Week, March 2l, l959.

'Chipman, J.S., "PERT With Costs," Technical Report ll2 SRP, WSPACS

Working Paper No. 4, Aerojet General Corporation, February l5, l96l.

-l3-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
requires that both cost and time be estimated with reasonable accuracy.

The pioneering effort for this case was accomplished by a team of en-

gineers and mathematicians representing the E. I. duPont de Nemours and

Co., Inc. and the Sperry-Rand Corporation. Their method was designated

as a Project Planning and Scheduling System* and was successfully

applied to a number of large and complex jobs involving design, con-

struction, and maintenance work. Two of the individuals largely res-

ponsible for this effort later joined the firm of Mauchly Associates,

Inc., Ambler, Pennsylvania. The Mauchly organization has offered

numerous workshop courses in many locations in the United States and

Canada to acquaint industry with the application of CPM. Construction

companies have been included in the groups participating in these

workshops and, in turn, have applied the techniques to their own pro-

blems .

Other work has been in progress by the Remington Rand, General

Electric, and IBM computer organizations. Each of them has developed

standard programs for the solution of certain scheduling problems

and is working on more advanced problems. Several other organizations

are also working on new or alternate methods. These include Lockheed,

Aerojet-General, and, in the construction field, the Bechtel Corporation

and Kaiser Engineers. Other construction companies, including Utah

Construction and Mining, Fruin-Colnon, Perini Corporation, and Peter

Kiewit, are already applying CPM to their operations.

Walker, M.R. and Sayer, J.S., "Project Planning and Scheduling," Report

6959, E.I. duPont de Nemours and Co., Inc., Wilmington, Delaware, March

l959.

-l4-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
THE CPM APPROACH - GENERAL

The Critical Path Method may be subdivided into three distinct

phases. Phase I is the construction of a representation, or network

model, of the sequential relationships of the operations making up the

project. This phase is a prerequisite for each of the other two phases.

It has been classified as a separate phase because its accomplishment

provides a useful product even if no further steps are taken. The

chart produced offers a better understanding than a conventional bar

chart of those relationships that control the order of performance of

the various operations. This understanding is valuable both in pro-

ject planning and control. Computer solutions differ little from a

manual solution insofar as this phase is concerned. The network

formulation is essentially a long-hand chore. However, computer pro-

grams have been developed for the proper numbering of the network

elements and their renumbering when, subsequently, operations are

added or deleted.

Phase II develops information primarily useful for control

purposes. From it, a construction schedule may be presented that

gives more information than a conventional bar chart. This is the

phase that determines which operations are critical, i.e., which de-

termine the overall project duration, and how much float the remaining

operations have. While this type of information is most directly

useful in project control, it also can have an important use in plan-

ning. Knowledge of operation floats can be used as a basis for

schedule shifts to smooth out labor and equipment requirements.

Considerable current effort is underway to devise computer programs

that will perform this shifting in the optimum fashion to achieve man-

power leveling in accordance with stated restrictions. Such programs

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

might simultaneously consider many different labor classification (or

different pieces of equipment). Phase II information can be developed,

even for relatively complex projects, by manual methods within very

reasonable limits of effort. The computer method's chief advantages

are in the elimination of errors and in the ability to quickly revise

-l5-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
the results if input information is altered. Probably the great bulk of

applications of Critical Path Scheduling in the construction industry to

date have gone only as far as Phase II. The fact that Phase I must be

accomplished manually and that Phase II may be accomplished manually

without unreasonable effort, indicates that for these two phases alone,

noncomputer methods offer invaluable aid to the industry.

Phase III introduces the possibility of performance time variations

and also introduces cost data. The objective is to determine the opera-

tion scheduling that produces the least direct cost for a given project

duration or to determine both the project duration and corresponding

operation scheduling that produces the least overall project cost. This

phase is the most difficult one to solve by any method. It is the one

that has been least applied. However, it is the phase that offers the

greatest possibilities for cost reduction. It is felt that the manual

method to be presented in this report can be used to solve problems

having a moderate degree of complexity. Moreover it is felt, as indicated

earlier, that there are certain advantages to be gained by working such

problems manually at the present stage of CPM practice.

Phase II not only precedes Phase III but also follows it. After op-

timum durations for each operation have been determined by application

of Phase III, then the corresponding new start and finish dates and

operation floats would be determined by Phase II.

To illustrate the procedures presented in this report, a project

that has been divided into a number of operations will be used as an

example. While this project will have only l8 operations, the time

and network relationships will make it complex for its size. The

factors encountered in a much larger problem will be present. As

the project is shortened, a total of six different critical paths

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

through the network will be developed. The entire process of short-

ening the project, (Phase III), starting at the point of least project

direct cost and proceeding to the all-crash, maximum direct cost point,

will be discussed in some detail. This will serve to acquaint the user

with the complete mechanics of the system. It will also duplicate the

full scope of information developed by the computer approaches. A less

-l6-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
detailed discussion will then be presented to indicate how the Phase III

information might be developed by working in the opposite direction from

a beginning at the all-crash end of the project time-cost curve. Finally,

these two approaches will be combined to offer a third approach. This

approach commences with a conventionally determined schedule and cost

estimate and applies Phase III mechanics to improve this schedule until

the optimum schedule is reached.

17
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
PHASE I

To develop any sort of detailed schedule, a project must first be

broken down into the operations necessary for its completion. A repre-

sentation of this breakdown on paper may be called a "model". When this

model indicates the sequential relationships of these operations, it

assumes the appearance of a "network". Hence the term "network model"

will be used to describe the chart showing the project breakdown and

sequential relationships between the resulting operations. Proper

visualization and construction of this network model is the most basic

and important step and is probably the most difficult part of the

Critical Path Method. This is as true for computer-oriented procedures

as for manual procedures. Network formulation is essentially a long-

hand job in both cases.

The degree of project breakdown can be varied considerably, but

certain factors often affect the divisions made. These factors include

the following:

(l) The type of work and predominant labor classification re-

quired for its performance. Operations performed by

different types of crews, such as electrical, dirt moving,

or concrete are generally separated.

(2) The structural elements involved. In a building, column

construction may be grouped separately from wall construc-

tion or floor slab construction.

(3) Responsibility for the work. Operations of various sub-

contractors are separated from those of the general con-

tractor.

(4) Location of the work. Work performed at different times

or by different crews because it is at different locations

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

in the structure may be considered separately.

(5) Owner's breakdown for bidding or payment purposes. On a

unit-price job, or a lump sum job where a breakdown for

progress payment purposes has been made, it is often

desired and sometimes required that the construction

-l8-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
schedule include the same breakdown.

The operational breakdown for formulation of the CPM Phase I net-

work also depends upon a careful definition of the sequential relation-

ships among the operations. Some operations must precede others, some

must follow others, and some may be performed concurrently with others

or independently of them. But every operation in the network must have

a definite event to mark its possible commencement. This event may

either be the start of the project or the completion of some other op-

eration which directly precedes it. Since the completion of an opera-

tion must signal the start of some other related operation, it is not

possible to have overlapping, related operations as are frequently

indicated on conventional bar charts. Where such a condition exists,

the operations must be divided further. The earlier work must be so

subdivided that the portion that must be completed before later work

is commenced is made a separate operation. And the later work must

also be subdivided, if necessary, so that just that portion made pos-

sible by completion of the preceding operation is included in a

separate operation. To more clearly illustrate these requirements,

consider a project involving a concrete wall. Let the operation of

constructing the footing for the wall be represented by A. Let the

operation of constructing the wall be represented by B. A portion of

the project network could be represented, then, as follows:

0©

This indicates that Operation B follows the completion of Operation A.

In other words, the entire footing is built before the wall is built.

Suppose, however, that the wall is long and that after the first half is

built, the wall overlying the completed portion can be started. A con-

ventional bar chart would indicate this condition by overlapping the

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

bars representing the two operations. However, the CPM network would

require dividing Operation A into two new operations, C and D, and

Operation B into two new operations, E and F. The equivalent portion of

the network would be represented as:

Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
This diagram states graphically that after the first half of the footing

is built (Operation C), then the first half of the wall may be built

(Operation E), and the remainder of the footing may also be built(Opera-

tion D). Note that it was also necessary to subdivide former Operation

B, since the construction of the second half of the wall (Operation F)

could not follow directly after the new Operation C. Care must be exer-

cised in drawing the connecting lines between operations in the project

network. For example, the foregoing figure states that Operation F, the

building of the second half of the wall, not only requires completion of

the footing but also requires completion of the first half of the wall.

This might be the case if, for example, the same forms were to be used

as on the first half of the wall. However, if building the second half

of the wall depended only on completion of the footing, the line between

E and F should be deleted.

These operational subdivision requirements tend to make the project

breakdown for CPM more complex than for conventional scheduling. However,

because the planner is forced to pinpoint more precisely the operational

relationships, a more valuable representation of the job is obtained in

the resulting network chart. It was mentioned that the degree of break-

down is variable. The operations chosen may represent relatively large

segments of the overall project or very detailed steps. For example,

the building of both footing and wall might have been lumped together

in the preceding illustration. On the other hand, the work might have

been broken down into considerably more detail to include such opera-

tions as: build forms, erect forms, place reinforcing steel, place

concrete, cure, strip forms, etc.

The project network may be represented in at least two ways. The

method generally used, and that on which the available computer programs
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

are based, involves an arrow notation. The method on which this report

is based involves the circle and connecting line notation just illustrated-

in the wall and footing example. One system of representation is easily

convertible to the other. The circle notation has been devised for the

manual method because it simplifies the mechanics of the Phase III pro-

cedure. It also simplifies construction of the network model. However,

-20-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
the arrow notation will be discussed briefly since it is widely used.

Moreover, since a different and useful form of progress schedule can be

presented in a modified version of the arrow chart, the use of both

systems might be desirable for those using the manual method.

In the circle and connecting line notation, each operation is repre-

sented by a circled number and its relationships to those operations that

either must directly precede or follow it are shown by connecting lines

to their circled numbers. The position of the circles in the network

representation is free to vary in the vertical direction. It is restrict

ed in the horizontal direction by the requirement that the horizontal

projection of a connecting line to any following operation must run from

left to right. One other restriction on horizontal positioning, though

not essential, produces a more orderly arrangement that is convenient for

data groupings in Phase III. A length for the shortest horizontal projec

tlon of a connecting line between two operations is chosen and is called

a "sequence step". The optional requirement, then, is that the circle

for every operation will be positioned horizontally the least possible

number of sequence steps from the left border of the network chart.

vThefirst step in construcJ^ing_a_ network model is to subdivide the

project into operations. Then a rough diagram is commenced, (see Fig.l)

There is no attempt to number the operations at this point. Rather,

each is represented by a circle with a descriptive title written next

to it. As each operation is entered on the worksheet, the questions

are asked:

(l) What operations must be completed to make this operation

possible?

(2) Can this entire operation be performed upon completion of

the preceding operation?

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

Answers to these two questions will often indicate that the prerequisite

operation must be broken down further or that the operation being entered

must be broken down further. Sometimes new operations are required. For

example, procurement of materials or delivery of equipment that must be

ordered in advance should be included as operations in the network.

Connecting lines are drawn between operations as the network is

-21-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
developed to indicate which operation must directly precede or follow

other operations. On this rough diagram the horizontal projection of a

connecting line from a preceding to a following operation may run in

either direction. It is therefore necessary to use arrowheads temporarily

to indicate the direction. When the rough network has been completed, if

it is desired to use the sequence step method for a more orderly final

arrangement, the sequence step of each operation can be determined. The

initial operations are given a sequence step number of zero. Each other

operation is given a sequence step number one greater than the highest

sequence step number of the operations directly preceding it.

The final network is drawn from the rough diagram, (see Fig. 2).

Each operation, indicated by a circle, is located horizontally to the

right of operations preceding it. Preferably, each operation is located

horizontally at its proper sequence step. Operations are located verti-

cally in such a position that connecting lines may be represented clearly

and, if practical, as straight lines. All connecting lines are drawn

but arrows heads are no longer necessary. The brief description of each

operation is entered next to the appropriate circle. Then all operations

are numbered by inserting the number inside the operation circle.

Numbering is commenced at the upper left corner of the network and precedes

vertically downward for all operations in the same column or sequence step.

Then numbering is carried to the top of the next column from where it

precedes downward in that column. This procedure continues until the ex-

treme right border of the network has been reached and all operations

have been numbered. The essential requirement for the numbering system

is that no operation will follow another operation of higher number. The

sequence step system is helpful in drawing and numbering the final net-

work. Once the maximum number of sequence steps has been determined on
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

the rough diagram, a sequence step can be assigned a horizontal dimension

that will make the network the desired length for its final portrayal.

Figure 1 shows a rough diagram for a simplified version of a small

pier job. The operations have not been numbered at this point. No

special attempt was made to draw connecting lines straight. Their direc-

tions are indicated by arrowheads. Sequence step numbers, the small

-22-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
numbers beside the circles, were determined upon completion of the net-

work. Figure 2 shows a final representation of this same network. The

operations have been plotted according to sequence step and have been

numbered. The operation descriptions are included for information

purposes but nil further computational work would be based on the as-

signed operation numbers. Figure 3 shows another and somewhat similar

project network. Although containing fewer operations, for convenience

in presenting the various worksheets that follow, its interrelationships

and data offer a considerably more complex problem than that for the

project of Figure 2.

The alternate, and more common, method of network representation is

the arrow diagram. In this method, each operation is represented by an

arrow whose tall is positioned at the point where the heads of arrows

representing directly preceding operations meet. Its head, in turn,

determines the point where the tails of arrows representing directly

following operations are to be positioned. Ordinarily, the arrow is not

a "vector" since its length and direction have no significance. However

the final network chart is drawn in such a manner that the horizontal

projection of each arrow, from tail to head, is a line running from left

to right. Moreover, for purposes of a new type of progress chart, the

length of each arrow may be chosen to make its horizontal projection

correspond to a selected time scale.

The arrow network is labeled to give the tail and head of every

arrow a number. The number of the head must always be greater than the

number of the tail. Arrows meeting and departing from a common point

have one number in common. Each arrow, and hence, each operation, has

two numbers that define it. The first number is that appearing at the

arrow tail and the second number is that appearing at the arrow head.
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

This method has the advantage, for computer applications, that the

resulting operation numbers indicate sequential relationships. For

example, every operation having a second number of 8 directly precedes

all operations witl' first numbers of 8. However, sequential relation-

ships in the manual method are determined by visual observation of the

network model. Therefore the simpler single numbering system is

-23-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
satisfactory. The arrow notation method has a disadvantage in the re-

quirement that all arrowheads of immediately preceding operations meet

at a common point and that all arrowtails of immediately following op-

erations commence at this common point. This is sometimes impossible

because a group of preceding operations may have one common following

operation but other following operations may not be common to the group,

or vice versa. This diagramming problem is overcome by introducing

"dummy" operations having zero time and cost.

Figure 4 is an arrow representation of the same network shown with

circle notation in Figure 3. As an example of the difference in numer-

ical designations of the same operation by the two systems, Operation

9 of Figure 3 becomes Operation 4-8 in Figure 4. (The operation numbers

from Figure 3 have been shown in Figure 4 in lieu of operation titles.)

With the arrow notation network it was necessary to add nine dummy op-

erations for this example. As an example to illustrate the necessity

for one of these dummy operations, consider first a relationship shown

in Figure 3. It is indicated there that Operations 8, 9, and l0 all

follow Operation 6 and that Operation l0 is preceded by Operation 2.

When we translate these sequential relationships to arrow notation on

Figure 4, we have no trouble in showing Operation 8 (now Op. 4-5) or

Operation 9 (now Op. 4-8) following Operation 6 (now Op. 2-4). But we

cannot show Operation l0 starting at the arrowhead of Operation 6, since

Operation l0 also follows Operation 2, but Operation 2 does not precede

either Operation 8 or Operation 9. The problem is solved by introducing

dummy Operation designated as Operation 4-6. All network relation-

ships are satisfied now because the dummy operation has zero elapsed time.

Operation l0 (now Op. 6-7) still follows Operation 6 directly even though

their arrows have been physically separated on the diagram by a dummy arrow.
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

The preceding paragraphs have discussed the arrow diagram and indi-

cated how it is constructed and how a circle diagram may be translated to

a corresponding arrow diagram. The circle notation will be employed in

presenting this report, except that the arrow diagram will be referred to

again in the discussion of Phase II, where its application to progress

charts is considered.

-24-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
PHASE II

The project now has been subdivided into operations and a network

chart prepared that shows all the operation sequencing interrelationships.

The only further information required for Phase II scheduling work is a

time estimate for each operation. These estimated times usually stem

from one of three situations, as follows:

(l) If the CPM technique is only to be applied through Phases

I and II then the estimated times represent some feasible

combination of operation times that permits completion of

the project in the allotted or desired time. These op-

eration times are determined by conventional estimating and

scheduling procedures.

(2) If Phase III of the CPM technique is also to be applied,

then the estimated operation times are those that give

the project duration and cost that serves as the starting

point for the Phase III computations. If Phase III com-

putations, follow the usual pattern for computer solutions,

this starting point would be the schedule that gives the

least direct project cost. The time for each operation

would be that at which it could be performed at least di-

rect cost. These times and costs are commonly referred

to as "normal time" and "normal cost."

With the manual method Phase III computations also

can be started at the project "all-crash" condition.

Then each operation would be scheduled at the shortest

possible, or crash, time. Later it will also be shown

how Phase III computations may commence with a conven-

tional estimate. Then the operation times would be the

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

same as described in (l) above.

(3) After Phase III of the CPM technique has been completed,

or carried as far as the planner wishes to take it, the

resulting revised operation times may then be fed back

into Phase II.

-25-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
With estimated times available for every operation, the computation

work of Phase II is performed on one worksheet. It is shown as Figure 5.

To simplify following the calculations, selected columns of figures are pre-

sented on four separate sheets. For actual computation work, a single sheet

would suffice. The eighteen operations are listed by number and the estimated

time of each is given in days. Since some of the results will subsequently

be used to illustrate Phase III computations starting at the project least

cost point, the operation times chosen are the normal, or least cost, times.

The procedures are exactly the same regardless of which basis is chosen for

operation times.

In job scheduling there is another choice to be made in connection with

the time basis. Time may be in working days or it may be in calender days.

The times given in the computations in this report will be assumed to be in

working days unless otherwise indicated. Care must be exercised to be con-

sistent. Such items as material or equipment deliveries are generally

quoted in calendar days, and the corresponding operation times must be con-

verted to working days if other operation times are in working days.

To perform Phase II computations Figures 3 and 5 must be used together.

Figure 5A shows the first step. For every operation, the earliest possible

starting date is determined. To this is added the operation duration time

to obtain the earliest possible finish date for the operation. The network

chart, Figure 3, shows that Operations l, 2, 3, and 4 may all start when

the decision has been reached to begin the project. Therefore zeros are

entered for their earliest start time. "Time zero" is the end of the day

before construction begins or actually the beginning of the first day. Op-

eration l, which requires 5 working days, has its earliest possible fin-

ish at time r>. This may similarly be interpreted to mean the end of the

fifth working day or the beginning of the sixth day. All other earliest
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

start and finish dates may be obtained from those of preceding operations.

For example, the network indicates that Operation 5 follows Operation l.

Therefore the earliest start date for Operation 5 is equal to the earliest

finish date of Operation l. If an operation follows more than one other

operation its earliest start date is equal to the latest of the "earliest

f1nish" dates of the preceding operation. For example, Operation l2 follows

-26-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Operations 5, 9, and l0. The network relations require that all three of

these operations must be completed before Operation l2 may be commenced.

Their earliest completion dates are, respectively, l7, l5, and 25. There

fore, the earliest start date for Operation l2 is 25. If operations are

numbered by sequence step groups as described in Phase I, it will always

be possible to run directly down the list of operations in numerical

order and determine earliest start and finish dates from preceding data.

The earliest finish date for the final operation is the total project

time. In this example the earliest finish of the final operation, Oper-

ation l8, is the end of the 64th day. This means that the project could

be completed in 64 working days.

'fy The next step of Phase II is to fill in the columns for the latest

start and finish dates for each operation. Figure 5B illustrates these

computations which are quite similar to, but the reverse of, those of

Figure 5A. A starting point is determined by using the earliest finis_h

time of the final operation as the latest finish time for the same oper-

ation. Then the latest start time is obtained by subtracting the opera-

tion duration time. From the bottom of the columns, computations are

carried upward. For each operation the latest finish time is equal to

the latest start of the following operation. For example, the latest

finish time of Operation l6 is equal to 6l, the latest start time of

Operation l8. Where an operation precedes more than one other operation

its latest finish time is equal to the earliest of the "latest start"

times of the following operations. For example, the network indicates

that Operation l0 precedes Operations l2, l3, and l4. The latest start

times of those operations have already been determined to be, respective-

ly, 35, 30, and 28. Therefore, the latest finish time of Operation l0

is 28.
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

Having determined both the earliest and latest possible finish dates

for each operation, it is possible to determine quantities known as

"Total floats." Total float, as shown on Figure 5C, is the difference

between the earliest finish time and the latest finish time. The

same results would be obtained by subtracting the earliest start time

from the latest start time. Float is a measure of the time leeway

ivailable for an operation. The total float figure states the number

-27-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
of days by which the finish date of an operation can exceed the earliest

possible finish date without affecting the duration of the overall project.

A zero total float indicates that an operation has no leeway and,

therefore, is one of the operations that establishs the project duration.

If its finish date were later, the project finish date would be later

by the same amount. Such operations are labeled "criticaI operations."

There must be at least one chain of critical operations running from

beginning to end of the project. Its total time gives the total pro-

ject duration time. There may be more than a single chain. If so, these

may be determined by.visual observation ot the network chart. To aid

in this process, those operations for which a zero total float has been

calculated may be double circled, or circled in color, on '-he project

network chart. Connecting lines between doubled circled operation

numbers may be doubled or traced in color. The resulting chains of

operations, or "critical paths," will be evident by observation. In

this example, there is a .single critical path consisting of Operations

3, l3, l5, l7, and l8. The sum of the operations times for these five

operations determines the project duration of 64 days. Should any of

these operations take more than the estimated time or be delayed at all,

the project time will be lengthened by a corresponding time. All of

the other operations have some degree of leeway. Any one of these may

be performed more slowly or have elapsed time between its start and

the finish of a preceding operation, as long as the increase falls

within the limits given by its total float. Within these limits, the

schedule change will not affect the 64-working-day project duration.

Additional scheduling information, known as "free float" time, can

also be obtained for each operation as indicated on Figure 5D. The

free float of an operation is the difference in days between its earl-

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

iest finish date and the earliest of the "earliest start" dates of all

of its directly following operations. With the aid of project network

chart observations the free floats can be calculated from data on

Figure 5D. For example, Operation 4 is shown on the project network

to be followed only by Operation l4. The earliest finish date of

Operation 4 is 20 and the earliest start date of Operation l4 is 25.

Therefore, Operation 4 has a free float of 5 days. As a second example,

consider Operation 6 which is followed by Operations 8, 9, and l0.

-28-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
The earliest finish of Operation 6 is ll. The earliest starts of the

three following operations are, respectively, ll, l1, and l5. The

earliest of these is the same as the earliest finish of Operation 6

which, therefore, has zero free float. The free float figure states the

number of days by which the finish date of an operation can exceed the

earliest possible finish date without affecting any other operation.

The concepts of "total" and "free" float can be illustrated graph-

ically by an adaptation of the common bar chart. Figure 6 shows such

a construction schedule for the example worked. The solid black bars

show operation durations and earliest start and finish dates. These

alone would be the equivalent of a conventional bar chart for the same

operational breakdown. Where there is no extension of the solid black

bar, there is no float, and the operation is a critical one. Extensions

of the bars represent the amounts of total float. However, the labels

"Interfering Float" and "Free Float" have been assigned instead. Free

float, as previously explained, is the time range over which variations

in operation completion time may occur without having any effect on

either project completion date or the amount of float of other opera-

tions. The difference between total float and free float is called

interfering float to indicate that, while operation completion in this

time range does not affect project completion time, it does affect some

subsequent operations by decreasing their floats. If there is zero free

float, then interfering float equals total float. This bar chart can be

constructed to also show calendar days, to provide space for actual pro-

gress bars, and to indicate periodic percentage completion estimates,

as is done with conventional progress charts. Of greater importance,

however, is the additional information for control purposes. When op-

eration performance fails to coincide with the conventional solid black

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

bars of the chart, management can better evaluate the importance of the

deviation. The chart also indicates possible scheduling shifts for plan-

ning purposes that will smooth out labor and equipment requirements or

improve financial demands. For example, any operation can be intention-

ally shifted within its free float range with no adverse effect on the

remainder of the schedule. Beyond that, operations can be intentionally

shifted within their interfering float range with a determinable effect

-29-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
on the amount of leeway of later operations.

Another form of construction schedule chart, utilizing the arrow no-

tation, is in use by industry. Figure 7 shows such a chart for the exam-

ple under discussion. Figure 7 is the network chart of Figure 4 with two

modifications. First, the arrows, except those for dummy operations, are

drawn to such length that their horizontal projections conform to a time

scale given both in working days and calendar days. Secondly, the arrows

representing the critical path for the network are plotted in a straight

line, horizontally through the center of the network, to give added em-

phasis. (Because dummy Operation D6 is required between critical Operations

l3 and l5 this network requires an offset in the straight line) Free floats

are indicated by extending the operation arrows to the appropriate junction

point with a wavy line. This type of chart has the advantage that it

retains "the network chart representation and so clearly indicates the

relationships between operations. It has the disadvantage that it

does not indicate directly the amounts of total or interfering float.

In summary, Phase II of CPM produces a valuable end product for

planning as well as control purposes. It may be applied to data avail-

able from ordinary estimating procedures provided that the project break-

down is made in a way that permits construction of the Phase I network

chart. Once the project network has been established and the operation

duration times estimated, Phase II calculations are purely mechanical.

There are no opportunities nor needs for judgment decisions during the

process. If a computer is available, this is an ideal use for it. The

results will be obtained very rapidly, and there will be fewer errors.

It is conceivable that the computer might be programmed to print out

a bar chart directly, with all the additional information available from

Phase II shown on it. If modifications are made in either the project

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

network or in the estimated duration times, the computer can produce

a revised schedule rapidly. On the other hand, because the amount of

effort required for a manual solution is not prohibitive it should be

clear that phase II of CPM will make a valuable planning and control

tool even when the services of electronic computers are not readily

avallable.

-30-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
PHASE III - GENERAL

Phase I of CPN provides a graphical representation of the sequential

relationships between the operations that make up a construction project.

Phase II provides valuable scheduling information for a single project dur-

ation. Neither of these phases introduce costs and, therefore, time-cost

relationships have not been considered. In the Introduction to this report

considerable emphasis was placed on the importance of a systematic method

for taking into account time-cost relationships and arriving at the most

favorable balance between them. Phase III does this and, for that rea-

son, is the heart of the Critical Path Method. The solution to CPM Phase

III provides the information for plotting a project time-cost curve and

for obtaining a complete set of operation lengths to correspond with all

points on that curve. The costs given by this curve are the lowest ob-

tainable total direct costs at the corresponding project duration times.

Since Phase III deals with time-cost variations, a starting point

from which to make these variations is the first prerequisite. One

point on the project time-cost curve that may be determined in advance,

with the aid of Phase II calculations, is that for least direct cost.

This is the starting point for computer approaches. The only other point

on the curve determinable in advance is that for "all-crash" performance.

It may serve as an alternate starting point. This report will propose

as a new starting point one of an infinite number of possible points

that lie above the ideal time-cost curve. These points are determined

by the project cost and duration as determined by a conventional estimate.

To establish the "least direct cost" starting point, each operation

is performed in the manner resulting in least direct cost. The sum of

the least costs for all of the operations gives the least possible

project direct cost. The corresponding operation durations can be

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

used as input data for Phase II and calculations can be performed as

shown on Figure 5A to determine the project duration that corresponds

to least project cost. These times and costs, for both operations and

the overall project, are designated as "normal time" and "normal cost."

Referring to Figure 8, point A represents a project normal time and

-3l-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
cost point. In the next section of thit- report, the mechanics for devel-

oping the curve AC, starting at point A, will be given.

Most operations can be speeded up from the normal time required for

performance at normal, or least, cost. This speeding up adds to the costs

as it represents such steps as overtime work, multiple-shift operations,

larger but less efficient crews, purchase of materials or services at

higher prices, and the use of more expensive equipment and methods.

Each operation has some practical time limit beyond which it cannot be

shortened further. This is referred to as the "crash" time and the cor-

responding cost is the "crash" cost. If the estimator provides crash

time and crash cost data for each operation, a second point on the pro-

ject time-cost curve may be located. This is the "all-crash" point

represented by point B on Figure 8. This represents the shortest possible

project time and may be obtained by using the operation crash times to

obtain the earliest finish date of the final operation by the procedure

Jof Figure 5A. The cost is the summation of the crash costs of every

operation and, therefore» is the highest possible cost. It would be,

for example, the cost to a contractor who, faced with a crash type job,

embarks on an across-the-board speed-up of all operations. However, gen-

erally the project crash time can be achieved at a cost much lower than

the all-crash cost. Many operations on the project crash time schedule

are not critical; i.e., they do not affect project duration. Hence, all

that is accomplished by crashing such operations is to increase project

cost without decreasing project time. The point on the project time-

cost curve that corresponds to the lowest cost for crash-time perform-

ance, represented by point C on Figure 8, cannot be determined until

Phase III computations are performed. A later section will give the

mechanics for developing the curve BCA, starting at point B.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

On Figure 8, point A represents the least project cost for the long-

est project time to be considered. Point C, undetermined as yet, repre-

sents the least project cost for the shortest possible project time.

There exist6 between these two points an entire range of possible project

durations. For each such duration there exists a schedule of operations

that gives the least project cost. Curve AC represents these optimum

-32-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
costs. To obtain curve AC, beginning at A, the procedure involves short-

ening critical operations that reduce the project time at the least in-

crease in project cost. This shortening is accomplished in steps using

in turn those operations or combinations of operations that have progres-

sively larger ratios of cost incurred to time saved. The curve from A

to C is not a smooth one as shown in Figure 8. Rather, it is a series of

straight line segments with increasingly steeper slopes. The result

is a concave upward curve that indicates that each new time reduction

comes at a higher cost.

Point C is reached when no funner project shortening can be ac-

complished. T~>i8 occurs when all t^e operations in one critical path

have been shortened to their crash lin.its. There may be operations in

other critical paths which can be shortened further, but it will be

found that such shortening makes them non-critical and does not affect

project duration. The critical operations at point C include not only

the original critical operations for the normal project schedule but,

generally, a number of additional operations that have become critical

as the project time has been shortened. As Indicated above, nothing

is gained after point C is reached. Non-critical operations, and crit-

ical operations that would become non-critical upon further shortening,

may still be shortened. However, this will not reduce project time

further, but will increase project cost as indicated by CB on the curve.

Another procedure for developing the projecc time-cost curve may

commence at the second of the two obtainable starting points on the

curve, point B. Briefly, this procedure requires first lengthening

all non-critical operations that are initially being performed at

crash rates. This lengthening permits cost decreases without length-

ening project time, as indicated by tne vertical line from B to C.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

After all non-critical operations have been lengthened either to their

normal time limit or until they have become critical operations, fur-

ther project cost reduction is achieved by lengthening critical oper-

ations and, hence, project time. In this case the operations, or

combinations of operations, that produce the greatest decrease in

cost per unit increase in time are first selected. A segmented curve

-33-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
from point C to point A with progressively flatter negative slopes results.

It would trace in reverse the curve from A to C obtained by the previously

described procedure. Point A is reached when all operations have been

lengthened to their normal times.

There are two reasons for developing this second alternate procedure.

The first is that the all-crash point of the project time-cost curve may

be considerably closer than the normal point to the desired solution.

This would be the case when an owner has established a contract comple-

tion date that obviously requires a near-crash effort to meet. Since

Phase III computations are often terminated when an optimum schedule for

the specified project duration has been obtained, it is advantageous to

start at the nearest point to this specified time. Secondly, the mechan-

ics of the procedure for moving to the right along the project time-cost

curve have been developed because ability to move in either direction is

required by the third procedure.

Certain other aspects of project scheduling variations can be shown

by Figure 8. An approximate reverse image of curve ACB can be produced

as curve ADB if, starting at point A, the least advantageous scheduling

changes are progressively made, rather than the most advantageous ones.

For example, if, first, all the noncritical operations were crashed

and then changes in combinations of critical operations producing the

greatest cost increase per day of project shortening were made, this

curve would result. The area enclosed by ACBD includes an infinite num-

ber of possible scheduling solutions for the project. For any given pro-

ject duration a scheduling solution represented by the point on curve

DB is the worst possible solution. And the scheduling solution represent-

ed by the point on curve AC is the best possib1e solution. In between

these are many, many, possible solutions of varying degrees of effective-

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

ness, any of which may be improved. A conventional estimate and schedule

would probably result in such a scheduling solution. Depending on the

skill and judgment of the estimator, the schedule might be represented

by a point very close to curve AC or some distance from it. Since this

distance, graphically and actually, represents dollars, a procedure to

improve such a schedule is of value. Once the best schedule for a given

project duration is determined, then further experimentation may be made

-34-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
in varying the project duration by movin.' either way along curve AC.

This discussion iudi >tes a third possible starting point tor the

Phase III procedure. This is represented hy point E on Figure 8 and

corresponds to the schedule and cost data obtained from a conventional

estimate. The first step is to lengthen any noncritical operations that

can be performed more economically with additional time. Curve portion

EF represents the cost improvement obtained In this manner. Then to

reach the ideal curve, whose position is as yet unknown to the planner,

a "wiggling-in" process must be employed. Alternately, the project is

lengthened over one cycle of computations by using the combination of

critical operations producing the steepest rate of cost decrease and

then shortened back to the original duration by using the combination of

critical operations that produces the least rate of cost increase. The

lengthening slope and the return slope tend to approach one another as

the process continues until a point is reached when they are the same.

This indicates that the ideal curve has been reached, and there is no

better possible solution for the given project duration. In Figure 8,

lengthening cycles FG, HI, and JK involved progressively less steep de-

creasing cost slopes. Return cycles GH, IJ, and KL involved progressive-

ly steeper increasing cost slopes. The final lengthening cycle retraced

along LK, the previous return cycle, and, therefore, signalled that

point L was on the optimum solution curve.

In order to perform the Phase III calculations, time-cost data are

required for each operation. Because the shortening or lengthening

cycles for developing the project time-cost curve generally involve

operation time changes over only a portion of their possible range, a

continuous time-cost function .for each operation is required. This in-

troduces the necessity for an approximation for two reasons. First, it

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

is only practical to require the estimator to give a limited amount of

data. Generally, the data would be confined to normal and crash times

and costs for each operation. Secondly, the methods developed for solv-

ing the problem, whether electronic computer oriented or manual, are

limited in their ability to handle more than the simplest type of time-

cost variations. The approximation generally adopted is that of a

-35-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
straight line variation of cost between normal and crash times. Perform-

ance of an operation in any intermediate length of time is accomplished

at a corresponding intermediate cost. This is a convenient and workable

assumption but some discussion of its reasonableness is necessary before

proceeding to the mechanics of the systems made possible by this approxi-

mation.

Figure 9 shows several approximations of operation time-cost relation-

ships. Figure 9A indicates the simplest one that has already been describ-

ed. This is the one commonly used for computer analysis and will be

employed in the noncomputer Phase III method that follows. Generally, it

is a sufficiently close approximation to make it consistent with other

estimating approximations. Figure 9B represents a more complex operation

time-cost variation. Its distinguishing features are that the estimator

has furnished more than two sets of time-cost data and that each resulting

segment slope, proceeding from right to left, is steeper. This type of

curve has the merit of often being a closer approximation to actual con-

ditions. Intuitively, it is recognized that for most situations it becomes

more and more expensive to obtain additional time savings. This operation

time-cost curve can be used in the Phase III calculations with little ad-

ditional effort. If a straight line approximation is unacceptable, this

extra effort is necessary. It is doubtful that this refinement would

frequently be justified for single operations, either from the standpoint

of the extra effort in applying CPM mechanics or in furnishing estimating

data. However, one special situation does justify its use. If a portion

of the project network can be replaced by a single operation without other-

wise altering any network relationship, this network portion can be treateo

as a separate "project." The resulting project time-cost curve could then

be used as the operation time-cost curve for the substituted single oper-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

ation. Such an operation curve would have the form shown by Figure 9B.

Another operation time-cost relationship is illustrated by Figure 9C

Its distinguishing features are that the estimator has furnished more

than two sets of time-cost data and that one or more of the resulting

segment slopes, proceeding from right to left, is less steep than its

preceding one. Here the segment GH might represent one method of perform-

-36-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
ance subject to variations by, say, overtime work or larger crew sizes.

Segment IJ might represent conversion to ar entirely different metnod

again subject to variations by overtime worn or changes in crew sizes.

The segment HI, which might be vertical, would indicate the additional

cost of one method over the other. This shape of time-cost curve cannot

be handled simply as was possible in the case of the Figure 9B curve. If

it can be handled at all by an electronic computer solution, only the

largest capacity computers now in existence would probably be able to do

so. The value of the additional accuracy compared to the cost of the

additional computational effort makes it doubtful if computer programs to

handle this type of cost-time curve will become available in the near fu-

ture. Fortunately, a straight line, such as GJ, eliminating the reverse

curve effect, generally offers an approximation thai is still reasonabl)

satisfactory. Figure 9D represents another situation that sometimes arises.

This is the case of a discontinuous operation time-cost curve where a

continuous straight line from K to N would not be a justified approximation.

Examples sometimes arise in materials procurement operations. To illus-

trate, consider Operation No. 3 of the network shown on Figure 2. This

operation involved the procuring of steel piling for use as pier fender

piles. For the lengths required by this West Coast job, piling could

only be purchased on the East Coast. The curve segment KL of Figure 9D

might represent several quotations and modes of shipment. Time might

vary from 60-75 days. An alternative was to obtain shorter sections lo-

cally and butt weld them to required lengths. The costs were considerably

higher but the time range was from l0-l5 days. Segment MN might represent

this possibility. Between l5 days and 60 days there were no intermediate

solutions. The problem was solved in the case of the project of Figure 2

by two complete solutions of the project time-cost curve, one based on

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

using a time-cost curve for Operation 3 equivalent to segment KL and one

using a curve equivalent to segment MN. The two project time-cost curves

were superimposed, graphically, and the most favorable sections of each

used for scheduling decisions. Obviously, this procedure is not a very

satisfactory one if many such operations are involved. The manual approach

would rule it out to a much greater degree than a computer approach. The

-37-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
alternative is to use the continuous straight line approximation for the

operation and make adjustments at the end of the problem to eliminate any

inconsistencies that result.

The acceptability of the operation time-cost curve approximation is

one of three basic assumptions on which the Critical Path Method is founded.

These assumptions will be discussed in the section "Evaluation and Applica-

tions" near the end of this report.

-38-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
PHASE III - NORMAL START

The mechanics of a noncomputer solution for CPM Phase III will now

be presented using the same example as for Phases I and II. In this sec-

tion of the report the project normal time-cost point will serve as the

starting point. This is the usual case for computer solutions and corres-

ponds to point A of Figure 8. Required data will be the same as that for

computer programs and the results will include all information obtainable

by computer methods. The method is based on the assumption of linear

time-cost variations for each operation as indicated by Figure 9A, and

upon the assumption that a realistic project network chart has been de-

veloped in Phase I as indicated by Figure 3. A detailed description of

the required steps will be given, but the worksheets are, in most cases,

in final form as they appear at the end of the procedure. For the reader

interested in following the details of the method, it would be advisable

to prepare fresh forms and work through the problem as it is discussed.

This applies particularly to the worksheets shown on Figure ll, l2, l3,

and l6. Figures l4 and l5 have been shown in separate stages as they

appear after each necessary revision.

It should be pointed out here that there will be a strong temptation

for a busy person to bog down and quit on the first attempt to follow

this procedure. It is involved and takes time to grasp. However, it is

mechanical and repetitive. Once the analyst understands it, the solution

becomes almost automatic.

Figure l0 contains the necessary new data for Phase III calculations.

These include the operation costs at normal production rates (normal cost

equals least possible cost) and crash times and corresponding crash costs

for each operation. Normal times are brought forward from Figure 5A. The

differences between normal time and crash time and between crash cost
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

and normal cost for each operation are computed as shown. Then the cost

difference is divided by the time difference to obtain the cost per day

for shortening each operation. This value represents the slope of line

AB in FigureHJA". "The two operations which the estimator says cannot be

shortened have no slopes. At the extreme right side of Figure l0, but

-39
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
using the estimated crash times, the earliest finish time of the final

operation is computed in the same manner as was done using normal times

on Figure 5A. This step is not essential, but it does determine the

project crash time quickly and provides a check point for the project

shortening process. The following data sums up what is known about the

project time-cost curve at this stage:

Normal (least) cost ■ $52,500

Normal time ■ 64 days

All-crash cost * $62,630

Crash time = 49 days

To obtain data for the project time-cost curve between normal time

and crash time limits as well as corresponding operation schedules, re-

quires six worksheets. These are briefly described as follows:

Worksheet No. l (Figure ll): This is a summary sheet on

which data from other sheets is entered as developed. Com-

putations are performed by cycles, and each cycle produces

a new segment of the project time-cost curve. First, the

combination of critical operations having the least cost

slope is determined and entered. These operations are

shown in the columns labeled "Min. Op. Slope Comb." Next,

the shortening limitations for these operations are deter-

mined and entered. One such limitation is the days of pos-

sible shortening for each operation to reach crash time.

These limits are shown for their respective operations in

the columns labeled "Op. Poss. Shortening." The other

shortening limitation is determined by interaction effects

in the network that produce new relationships requiring

other worksheets to be updated. It is entered in the column

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

labeled "Interact. Limit." The governing limitation is

entered in the column "Days Change" and is used with the

cost slope to calculate a cost change for the cycle.

Finally, new project cost and time values are computed.

Then the next cycle is commenced.

Worksheet No. 2 (Figure l2): The left portion of this sheet

-40-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
is a tabulation of project operations arranged in order of

increasing cost slopes. Entries indicate whether an opera-

tion is critical or not by showing the cycle in which it

becomes critical in the column labeled "CP. Cycle." Entries

in the column "Finish Cycle" indicate the cycle in which a

critical operation has been shortened either to its crash

limit or to a point at which further shortening would make

it noncritical (NC). The foregoing information assists in

selecting the most favorable operations available for short-

ening. The right portion of this sheet is a tally record on

which entries indicate the revised operation times and the

remaining possible shortening as changes take place. The

original "Normal Time " and "Poss. Short." are entered from

Figure l0. The latest entry for an operation provides one

of the shortening limitations for the cycles of Worksheet

No. l. The entries also provide complete operation schedul-

ing information at the end of each shortening cycle.

Worksheet No. 3 (Figure l3): This sheet provides the other

shortening limitation for the cycles of Worksheet No. l. This

limitation is determined by changes in the network relation-

ships that require updating of Worksheets Nos. 4 and 5 and

possibly Worksheets Nos. 2 and 6. The latter sheets are

involved when the updating of Worksheet No. 5 indicates that

critical operations have been added or lost.

Worksheet No. 4 (Figure l4): This sheet is known as a pre-

cedence matrix^ It indicates all of the additional opera-

tions that are affected by a change in any specified opera-

tion jind provides this data for use on Worksheet No. 3.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

It must be updated at intervals as indicated by information

developed on Worksheet No. 3.

Worksheet No. 5 (Figure l5): This sheet is the project net-

work. It is used to determine the conversion of noncritical

operations to critical ones, or vice versa. Each connecting

line of the project network shows that one operation follows

another. The term "lag" has been adopted in this manual

-4l-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
method to indicate the days between the earliest completion

of an operation and the earliest start of a following opera-

tion. These lags for each connecting line are calculated on

Worksheet No. 3. When a lag is, or becomes, zero, the cor-

responding connecting line is doubled and shows that one op-

eration immediately follows another. When visual observation

i indicates a continuous chain of doubled connecting lines,

* a critical path has been formed. The corresponding connect-

/: ing lines are then tripled to show this path, whenever a

critical path is added or lost Worksheets Nos. 2 and 6 must

be updated.

Worksheet No. 6 (Figure l6): This sheet provides the com-

binations of operations having the least cost slope for

entry on Worksheet No. l. It is only required when there

is more than one critical path. The critical paths are

listed in pairs on the upper portion of the sheet. On

the lower portion a systematic trial and error procedure

determines the optimum combination of operations.

A more complete description of the use and development of these worksheets

will be included as procedural steps are discussed.

The starting point is cycle 0 on Figure ll. The normal project cost

and time are entered. Cycle l, the first shortening cycle, is next. The

first question is "Which operation, or operations, should be shortened

to produce project shortening at the least increase in project cost?"

Three factors determine the choice. The jy>eration_must,be_ critical; it

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

must be possible to shorten it within its crash limit; and, since there

is only one critical path at this time, it must be a single operation

that has the least possible cost-time slope. A selection might be made

from Figure l0, but it is much easier to first rearrange the operations

in increasing order of cost slopes as shown on Figure l2. The third

column on Figure l2 shows the shortening cycle at the end of which the

operation became critical. At the outset of shortening the only entries

in this column are the zeros opposite Operations Nos. 3, l3, l5, l7, and

l8. These are the critical operations as determined in Phase II on

-42-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Figure 5C and as also shown on the project network of Figure l5A. The

normal time, possible shortening, and revisions of these figures occupy

most of the remainder of Figure l2. The only remaining entries are in

the fourth column, and they indicate the cycle at the end of which the

operation can, or should, no longer be shortened. One basis for such

an entry is that the remaining possible shortening,has„reached^ero as

indicated on the tally portion of the sheet. The other basis is_ that

any further shortening will make the operation noncrltical. This is

determined by data on Figure l6 which will be discussed later. Examples

of such entries are those followed by the letters NC. Having develop-

ed Figure l2, it is simple to select the operation meeting the three

criteria given. It will be the first operation, starting at the top of

the tabulation, that has an entra in column 3 and^no entry in column ^.

For Cycle l, Operation 3 satisfies this requirement. (At this time there

are no entries in column 3 opposite Operation 5 or in column 4 opposite

Operation 3.)

Having selected the operation to be shortened, the next question

is how much can it be shortened? There are two limitations that deter-

mine the answer. )$ne is the amount of possible shortening of the oper-

ation, which can be obtained from the latest possible shortening time

entry on Figure l2. For Operation 3 at The beginning of Cycle l this

is l2 days. The second limitation is referred to as the interaction

limit. This_ limit represents a point at which network relationships

change andJuj>osslbly, a new critical path is introduced. Determination

of this limit is the most involved problem of the Phase III procedure.

Several methods have been explored and the one offered here appears to

be the most workable.

Figure l3 provides a worksheet for interaction limit determination.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

It is used in conjunction with a triangular matrix, known as a "Prece-

dence Matrix," and the project network chart. The first two columns

of Figure l3 give the operation numbers that determine, respectively,

the beginning and end of every connecting line in the project network.

Starting with Operation l and referring to the network chart of Figure

3, there are two connecting lines originating at this operation, l to

5 and l to 6. Operation l is listed as the preceding operation, or

-A3-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Pre. Op., in column I and Operations 5 and 6 are listed as the fo1lowing

operations, or "Post Op.", in column 2. This process iB continued until

every operation appears in column l, in numerical order, and every con-

necting line in the network is represented by a column l - column 2 pair

of numbers. Next, in column 3, the earliest finish time, from Figure 5A

is entered for the operation appearing in column l. In column 4, the

earliest start time, from Figure 5A, is entered for the operation appear-

ing in column 2. The differences between the earliest start times of

the following operations and the earliest finish times of the preceding

operation are entered in column 5 under the heading of "Lag." Our results

show, for example, that between Operations l and 5 there is zero lag.

This indicates that when each operation is performed at its earliest

possible time, Operation 5 commences immediately upon the completion of

Operation 1. Between Operations 4 and l4, however, there is a lag of

five days, indicating that Operation l4 does not start until five days

after the completion of Operation 4. The word "lag" has been introduced

here instead of the word "slack", since "slack" has become generally used

in the probabilistic scheduling cases as the counterpart of the word

"float" in the deterministic cases.

From the data developed so far on Figure l3, the precedence matrix

can be constructed. On a square grid system, shown on Figure l4A, the

operation numbers for all operations are listed vertically along the

left border and horizontally across the top. The numbers on the left

border represent preceding operations for network connecting lines,

while those along the top represent following operations for network con-

necting lines. The zero lag relationships developed on Figure l3 are

entered on this matrix by entering zeros in the proper grid boxes which

represent the preceding operation and following operation of all con-

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

netting lines having zero lag. No entries are made where Figure l3

indicates a positive lag. For example, Figure l3 indicates that Opera*

tions 8, 9, and l0 each follow Operation 6 with lags, respectively, of

0, 0, and 4 days. This information is transferred to the precedence

matrix by entering on the row of Preceding Operation 6 a zero in the

column of Post Operation 8 and a zero in the column of Post Operation

9. No entry is made in the column of Post Ope ration 10. To complete

-44-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
the first stage of the precedence matrix, shown as Figure l4A,,jme_jentrx.

lsjmade. for every connecting line in the network that has zero lagas

tabulated on Figure l3. A diagonal line bisecting the square grid is

drawn for use in the following stages of precedence matrix development

and updating. The information shown separately on Figures l4B to l4G

would normally be developed on the same original sheet as Figure l4A The

procedure is shown in separate stages in this report merely to facilitate

the explanation.

On Figure l4A the precedence matrix indicates only those operations

which immediately follow other operations. On Figure l4B the precedence

matrix is developed further to indicate, by the addition of x entries,

the entire chain of operations, that ,i.nmiediately follow one another, time-

w^se^ starring with the operation o£_the row In which.the entries are

made.. For example, Figure l4B indicates that any time variation in Op-

eration l3 would affect Operations l5, l7, and l8. For a graphical

illustration, visualize a bar chart representing the operations of this

project. If Operation l3 were shortened by one day, the bar represent-

ing Operation l3 would be shortened by one day, and the entire bars

representing Operations l5, l7, and l8 would be shifted one day to the

left. (An exception to this statement will be explained below.)

The x entries on the precedence matrix are obtained by starting in

the last row of the grid which contains a zero. Commencing at the

intersection of the diagonal line with that row, move vertically upward

in the column containing this intersection to squares containing zero

entries. When_zero ejit_rAejS_are._e.ncountered, place an x in their rows

directly above—the zero entries (and in subsequent steps also above the

x entries) of the starting row. To illustrate this procedure with re-

ference to Figure l4B, the last row containing a zero is that of Opera-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

tion l7. This zero states that Operation l8 immediately follows Opera-

tion l7. Starting where the diagonal line intersects this row, proceed

up column l7 to the only zero encountered, which is in the row of Oper-

ation l5. This zero states that Operation l7 immediately follows

Operation l5. An x is entered in this row in column l8, directly above

the zero in row l7. This x states that Operation l8 is directly affected

-45-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
by any change in Operation l5. The reason is that Operation l8 immediate-

ly follows Operation l7 which immediately follows Operation l5. Having

completed the entries originating from the bottom row, return to the next

row up containing a zero and repeat the same procedure. This would be

row l5. Proceeding up column l5, the only zero encountered is in row

l3. X's are entered in row l3 directly above the zero and x entries in

row l5. (The entries in row l3 now give the information discussed in

the example of the preceding paragraph.) This procedure is repeated a

row at a time until the top of the grid is reached.

There must be at least one zero in every column of the precedence

matrix (exceptthosei representing initial operations), because every

operation can immediately follow at least one other operation. There

may be more than one zero in a column. This raises a detailed point

that it is necessary to discuss. Suppose there were another zero in

column l5 in row l2 (a possibility since connecting line l2-l5 has a

positive lag subject to change as shortening takes place). It was pre-

viously stated that shortening Operation l3 by one day would cause the

bar representing Operation l5 on a bar chart to shift left by one day.

This is another way of saying that both the earliest start time and

earliest finish time of Operation l5 would be reduced by one day. How-

ever, if Operation l5 also directly follows Operation l2, its bar could

not shift unless Operation l2 either was shortened or shifted also.

The mechanics of the system automatically take care of this problem.

There are two possibilities. If Operation l5 is critical, then both

Operations l2 and l3 would be critical since they immediately precede

Operation l5. Therefore, Operation l3 would not be shortened unless

either Operation l2 or some operation in its critical chain were short-

ened concurrently. Therefore, Operation l5 could shift as stated.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

However, if Operation l5 were not critical, Operation l2 would not neces-

sarily be critical. If it isn't, shortening Operation l3 would not

require an accompanying change in Operation l2. Therefore, if the fin-

ish time of Operation l2 did not change, the start and finish times of

Operation l5 could not change. Rather, connecting line l3-l5 would

change from a zero lag to a positive lag. This would be indicated on

-46-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Figure l3 for the cycle involved, And the precedence matrix would have

to be updated. This updating would require removing the zero in the

grid box for connecting line l3-l5 and also removing all x's from the

matrix that have been entered solely as a result of that zero. A more

accurate statement than that made earlier concerning the relations

shown by the precedence matrix can now be offered. The zero and gen-

tries in any row of the precedence matrix indicate the following opera-_

tions for which scheduling times may be affected by a shortening of Jtjje__

operation given by the row number. Whether these operations are actually

affected is determined by one cycle of entries on the Network Interaction

Limit (Figure l3) worksheet. If a zero lag becomes a positive lag, one

or more entries on,theprecedence matrix will be removed^. The scheduling

times for the corresponding operations will not have been affected by

the operation shortened.

Note that the preceding statement was confined to operation short-

ening. If operation lengthening is involved the corresponding state-

roent becomes simply: The zero and x entries in any row of the precedence

matrix indicate the following operations for which scheduling times will

be affected by a lengthening of the operation given by the row number.

Generally, contract change orders and extra work orders involve length-

ening of operations. It is convenient to refer to a row of the pre-

cedence matrix and state definitely which operations will be affected.

It is still advisable to work through Network Interaction Limit cycles

for the lengthening time required, because some positive lags may be-

come zero causing additional entries on the precedence matrix.

Having developed the initial precedence matrix on Figure l4B, the

interaction limit for the first cycle may be determined on Figure l3.

Under the cycle l heading and on the rows opposite the appropriate
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

"Pre. Ops.", entries are made in the left half of the column to indicate

which Pre. Ops. have their earliest finish date shifted to an earlier

time by the shortening of Operation 3. There are two bases for these

entries. One is that the operation beinp shortened wi ll have_ its

earliest finish date shifted. The second is that all operations witti

a zero or x on the precedence matrix In the row of the shortened op-

-47-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
eration will have their earliest finish dates shifted. These two criter-

ia indicate that entries should be made opposite Pre. Ops. 3, l3, 15, and

l7. (There is no Pre-Op. l8.) Next, a simijar^enjEry is made in the

right half of the column to indicate which Post Ops. have their earliest

start dates shifted. There is only one basis for these entries. It

is the information from the precedence matrix indicating, again, the

same operations affected by a change in the shortened operation. These

are Operations l3, l5, l7, and l8. These entries are less convenient

to make since the Post Ops. are not listed in numerical order. For each

operation that is affected as shown by the precedence matrix, a run must

be made down the Post Op. column. A valuable check on the work can be

maintained if the network chart is referred to when making the entries

for each operation. The number of connecting lines terminating in an

operation on the chart indicates the number of entries that should be

made for that operation in the right half of the column. For example,

the chart of Figure 3 (or l5A) indicates that three connecting lines

terminate at Operation l7. Therefore, there are three entries to be

made. A careful run down the Post Op. column will show that l7 appears

three times. Such a check will eliminate the possibility of omitting

entries.

The horizontal V symbols in the cycle l column of Figure l3 indi-

cate all shifts in earliest start and finish dates. They are pointed

to the left in keeping with the bar chart concept of changes to earlier

dates. If there are no entries in a row this indicates that neither

the finish of the preceding operation nor the start of the following

operation will change, and, therefore, the lag will not change. Two

horizontal V's in a row indicates that both operations will shift, and

there will still be no change in the lag. A single horizontal V in

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

the left half of the column, for a shortening cycle, indicates that

the earliest finish of the preceding operation will shift to an earl-

ier date, and the lag will be increased. The fourth, and only other

possibility, is a single horizontal V in the right half of the column

which indicates that the earliest start of the following operation will

shift to an earlier date, and the lag will decrease. When a lag decreases

-48-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
from any positive value to zero, the network relationships are altered

and the precedence matrix will change. At this time an additional crit-

ical path may or may not be introduced. In the example used here, a new

critical path is added each time a lag becomes zero. This does not have

to be the case. But, in any event, when a lag becomes zero, changes take

place that require an interruption to update the data. By running down

the cycle l column and observing the lags for those connecting lines

for which the entries indicate that lags will be shortened, the least^

of these lags is determined. This is the "Interaction Limit." In the

example, it is equal to 2 days and is established by the lag of connect-

ing line ll-l7.

After the interaction limit has been determined for cycle l, it is

entered on the Summary Sheet, Figure ll. Entries for cycle l may now be

completed on this sheet. The number of days of shortening for a cycle is

determined either by the interaction limit or by the possible shortening

ojf the operations involved. Th» «mfl]J«mr number mint- be used In this

case, cycle l produces 2 days of shortening as controlled by the interac-

tion limit. Since Operation 3 has a cost slope of $50 per day, project

costs are raised by $l00 and project time is reduced to 62 days.

Before proceeding with cycle 2 the supporting data sheets oust be

updated. The actual number of days of shortening is entered on Figure

l3, and all of the lags for which cycle l entries indicated changes are re-

vised by this number. Only those that are changed need be entered in

the revision column. Whenever expositive lag becomes zero or a zero

lag becomesjpositivej, a_ revision tojthe precedence matrix must be made.

In this cycle, the lag between Operations ll and l7 was reduced from 2

to zero. The revised precedence matrix is shown as Figure l4C. Re-

visions are indicated by the bordered blocks. One comment is warranted.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

When a zero is entered in the square of row ll - column l7, it is

necessary to backtrack downward before the new data can be carried upward

on the matrix. To illustrate, starting at the newly entered zero,

drop down column l7 to the row of the diagonal line intersection. Then

carry up to row ll, as x's, any zero or x entries in this row. In this

case there is a single zero in the row of the diagonal intersection,

-49-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
row l7. An x must be placed in this column in row ll. Now proceed to

carry the new entries of row ll up the chart in the usual manner. The

precedence matrix is now ready for cycle 2.

It was mentioned that when a lag becomes zero a new critical path

may or may not be added. It is essential to determine whether or not

this has happened and, if so, to identify it. This can most easily be

done by visual means using the network chart. As in the case of the

precedence matrix, a single network chart would be used and updated

as necessary. To clarify this explanation, successive revisions to the

network chart are shown as Figures l5A to l5F. On the initial network

chart of Figure l5A, all connecting lines representing zero lag have

been drawn double. Where these doubled connecting lines form a chain

from start to finish of the project, they represent a critical path.

In this case they have been shown as triple lines to further distinguish

them. Other means, such as colors, might serve to classify their status.

On Figure l5A, the connecting lines between Operation 3, l3, l5, l7,

and l8 have been tripled to represent the original critical path

through the network. Additional critical paths need not form a

complete new chain from end to end of the project. They may only in-

volve new subchains between either end of the project and some opera-

tion on an existing critical path, or between intermediate points on

existing critical paths.

At the end of cycle l, the network chart would be updated by doubling

connecting line 1l-l7, the lag of which has become zero. This is shown

on Figure l5B. A visual observation indicates a new critical subchain

between the beginning of the project and Operation l7. This subchain con-

sists of Operations l, 5, 7, ll, and l7. The corresponding connecting

lines are tripled to indicate the second critical path.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

Figure l2 must also be updated before commencing cycle 2. A revised

possible shortening of l0 days is entered in the cycle l column opposite

Operation 3. While it is not essential to enter this information in a

separate column for each cycle, it adds convenience if the analyst wishes

at the completion of the problem to quickly pick off the entire operations

schedule as of the end of any particular cycle. It also facilitates check-

-50-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
ing the work. Since a new critical path has been developed, this is in-

dicated by entries in the third column opposite the new critical operations.

The "l" signifies that the operation became critical as a result of the

changes made during cycle l.

The fact that there is now more than one critical path introduces

an additional problem in selecting the operation, or operations, to short-

en for the next cycle. Where the jnritical paths are parallel, as they

are between the beginning of the project and Operation l7, shortening a

critical operation on one path necessitates shortening a critical opera-

tion on the other j>ath;_ otherwise the entire project would not be short-

ened. The alternative is fco shorten a single critical operation common

to both paths, such as Operations l7 and l8. The problem is to determine

the least costly choice, whether it is to shorten a combination of two

operations simultaneously or to shorten a single operation. While this

is not particularly difficult to do if there are only two critical paths,

the decision rapidly becomes more complex as additional critical paths

are added. Therefore a system must be established for making the de-

cision. Figure l6, used in conjunction with Figure l2, presents such

a sy^cem. For explanation purposes, Figure l6A presents the initial

version of this worksheet and Figure l6B presents the final version after

the last cycle of shortening is completed.

\/ At the top of Figure l6A under the heading of Cycle l, are shown the

two subchains, 3-l3-l5 and l-5-7-ll, that, as a result of cycle l shorten-

ing, have become simultaneously critical between the beginning of the

project and Operation l7. The operation in each path that has the least

cost slope is circled for quick reference. The operations in each path

are listed vertically according to their sequence steps. At the left of

the tabulation all possible operations in each sequence step are indicated
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

as an index. This index is not essential, but as more critical paths

are introduced or as projects with larger numbers of operations are

analyzed, this procedure makes it easier to locate the paths containing

a given operation. For example, if Operation l2 appeared in any critical

path, it would be in the row containing the ll to l4 sequence of operations.

For only two paths and a few operations it is just as easy to scan the

-5l-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
two columns for Operation l2.

On the bottom portion of Figure l6A all the operations subject to

change are listed, as on Figure l2, in order of increasing cost slopes.

The cost slopes are also relisted for convenience. The process of se-

lecting the most favorable combination of operations to be used for cycle

2 shortening is now commenced. The most eligible single operation,

according to our previously developed criteria for using Figure l2, is

Operation 5. An x is placed opposite Operation 5 on the bottom portion

of Figure l6A in the first column under the heading of cycle 2. A

check of the top portion of Figure l6A indicates that Operation 5 does

appear in one of the portions of a critical path that has a parallel

critical path. Therefore, nothing is gained by shortening Operation 5

alone. In additionJ_,the_least expensive operation in the parallel path

must also be shortened. The circled number indicates that this is

Operation 3, and an x is placed opposite it in the lower table. The

sum of the cost slopes of shortening these two operations, $90 per day,

is entered at the foot of the column. This is not necessarily the best

solution as there are other operations to try that may give a smaller

figure. The next eligible operation according to Figure l2 is Operation

3. This is one of the operations of the first combination and need not

be tried again, since the same results would be obtained. The next

eligible operation is Operation IS, which is entered on Figure l6A in

a new column under the cycle 2 heading. By repeating the previous pro-

cedure it is determined that Operation 5 would have to be concurrently

shortened. An entry is made accordingly, but no attempt is made to sum

the cost slopes. The fact that one of the two operations is the same

as that entered in the first column under the cycle 2 heading, while

the second operation falls further down the listing, indicates that
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

a higher sum would result. There is still no basis for discontinuing

further trials. Figure l2 indicates Operation l7 as the next eligible

candidate. After the entry is made on Figure l6A, a check of the upper

portion of the worksheet indicates that there is no parallel critical

path to that containing Operation l7. (Operation l7 is common to both

critical p;iths.) Therefore, the entire project can be shortened by

-52-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
shortening Operation l7 alone at a cost slope of $70 per day. There is

no need to proceed further, since Figure l2 indicates that the next

eligible operation for trial is Operation l3 at a cost slope of $l80 per

day. Operation l7 is entered on the Summary Sheet, Figure ll, as the

minimum slope selection for cycle 2. To determine how much Operation l7

may be shortened involved the entire procedure developed for cycle l.

There are no new concepts required for this cycle.

All of the steps and worksheets for solving the remainder of the

problem have been developed at this point. A third, fourth, fifth, and

sixth critical subpath is introduced, respectively, by cycles 2, 6, 7,

and 8. The least cost combination of operations to be concurrently

shortened requires a maximum of four operations in cycle 7. The.JJjnlt

of project shortening has been reached when there are no critical oper-

ations that may be shortened effectively_as shown by column 4 of Figure

l2. In the example, entries in this column indicate that every critical

operation has reached its shortening limit after nine cycles. This limit

should check the predetermined crash limit calculated at the outset of

the Phase III work.

In column 4 of Figure l2, several of the entries are for operations

which have not been shortened to their crash limits. These entries give

a cycle number followed by the letters "NC". This means that as a re-

sult of tjte_jchanges_produced in the jjiven cycle, any further shortening

pfthec^r^8£Ondlng. operation would cause it to be noncritical. The

basis for such an entry comes from the critical path comparisons in

the top portion of Figure l6B. When it becomes impossible to shorten

one of two parallel critical subpaths because all operations in the path

have been shortened to their crash limit, then the effect of shortening

an operation in the other parallel path is to make that path noncritical.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

The operations in the path would, therefore, become noncritical unless

they are also in other independent critical paths. These possibilities

are Illustrated by the example. At the end of cycle 8, subchalns 2-l0

and 3 become parallel critical subpaths between the beginning of the pro-

ject and Operation l3. However, Operation 3 had, in that same cycle, been

shortened to its crash limit. Any shortening of Operation 2 (Operation

-53-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
l0 had also previously been shortened to its crash limit) would cause

the subchaln 2-l0 to be noncritical between the project beginning and Op-

eration l3. However, it does not follow in this case that shortening of

Operation 2 would cause it to be noncritical since it is part of another

critical path that is independent of Operation l3. Therefore, no entry

was made in column 4 of Figure l2 at the end of cycle 8, and Operation

2 was still considered in arriving at the least slope combination for

cycle 9. Had Operation 2 been one of those shortened in cycle 9, the lag

of connecting line l0-l3 would have increased from zero to a positive

value, and that connecting line would no longer have been on a critical

path on the project network chart.

At the end of cycle 9, all operations in the critical subpath l-5-7-ll

have been shortened to their crash limits as indicated by the fact that

these numbers have been struck out on the tabulation under cycle l at the

top of Figure l6B. It follows that shortening either Operation l3 or IS

would make the parallel critical subpath, 3-l3-l5, noncritical between

the project beginning and Operation l7. Since this is the only critical

path containing Operations l3 and l5, it can be said that they would be-

come noncritical operations if shortened further. This justifies the

entries for these operations in column 4 of Figure l2. Moving to the

next pair of critical paths on Figure l6B, a similar justification may

be made for Operations 2 and l4. While Operation 2 is common to two in-

dependent critical paths, both of these paths would now become noncritical

if Operation 2 were shortened. In a similar fashion, the remaining pairs

of critical subpaths give justification for the entries in column 4 of

Figure l2 for Operations 4 and 6. If, by oversight, an operation that

would become noncritical upon shortening were used for a shortening

cycle, an interaction limit of zero would result on the Figure 13 work-

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

sheet. This would demand an investigation which should point out the

oversight.

To sum up the mechanics of performing the Phase III calculations,

a list of ten procedural steps for each cycle follows:

l. Determine the critical operation, or combination of

critical operations with least cost slope. Use Figure

-54-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
l2. If there is more than one critical path, also use

Figure l6.

2. Enter the selected operations opposite the shortening

cycle number on Figure ll. Also enter the possible

shortening for each operation, and the cost slope of

the combination. Data for this comes from Figures l2

and l6.

3. Enter operations to be shortened on Figure l3 and proceed

to determine the interaction limit with the aid of the

precedence matrix, Figure l4.

4. Enter the resulting interaction limit on Figure ll, de-

termine the days the project is to be shortened, and

complete the cycle data, thus obtaining a new Total Pro-

ject Cost and Total Project Days.

5. Enter the days shortened on Figure l3 and revise all lags

that change during the cycle.

6. If any positive lag has become zero or any zero lag has

become positive, update the precedeice matrix, Figure l4.

7. If any positive lag has become zero or any zero lag has

become positive, update the project network chart, Fig-

ure l5.

8. If the project network chart indicates a new critical

path (or loss of a critical path), update column 3 of

Figure l2 and the upper portion of Fig re l6.

9. Update the tally portion of Figure l2. If any opera-

tion has reached its crash limit, update column 4 of

Figure l2.

l0. If any entries have been made in column 4 of Figure l2,

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

check the top portion of Figure l6 and strike out the

corresponding operations. If a circled operation is

struck out, circle another operation in the same column

with the next least cost slope. If all operations in

a column have been struck out, che:k to see whether fur-

ther entries for noncritical shortening limitations

-55-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
should be made in column 4 of Figure l2. If such entries

are made, repeat this step.

Figure l7 illustrates graphically the information from the Summary

Sheet, Figure ll, for the entire project shortening curve. From this

graph, for any given completion time, the least total project direct cost

can be determined. Moreover, the scheduling for every operation necessary

to meet this project time and cost may also be determined. The schedul-

ing is obtained by deducting from the normal time of each operation the

number of days of shortening for those operations indicated by the graph

between the project normal time and the specified time. A more direct

approach would be to refer to the tally portion of Figure l2. The com-

plete schedule at the end of the cycle in which the specified project

completion time falls is available. Those operations shortened during

that cycle would be increased by the number of days between the specifi-

ed project time and project time at the end of the cycle.

If the procedure of Phase III is carried from the project normal

time to the project crash time, the full range of optimum schedules and

project costs for any possible completion time is available. Remember-

ing that only direct costs have been considered, one use of the project

time-cost curve is to combine it with other time-cost data relating to

project performance. From a contractor's viewpoint, a second curve rep-

resenting indirect costs could be developed. In those cases where there

were bonus-penalty provisions, a third curve could be added representing

this incentive for completion earlier than the specified time. All

curves would be plotted on a common basis of either working days or cal-

endar days. A summation of the curves would result in a project time-

total cost curve having a minumum point. The schedule corresponding to

this minimum point would be the optimum schedule since it would give the
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

lowest total project cost.

It is not essential in many cases that the entire project time-cost

curve be developed. Initially, the contractor might carry the calcula-

tions only far enough to obtain project compression to the point of

specified completion time. This would give him the best schedule to meet

the project duration requirements of the owner. A further refinement

-56-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
from that point might then be made by considering the indirect cost curve,

as described above. It still might not be necessary to develop the entire

project time-cost curve. A slight additional projection of it, added to

a corresponding segment of the indirect cost curve would usually indicate

the minimum point of the combined curves. The entire indirect costs need

not be developed to plot the indirect cost curve. Only the variable por-

tion of the indirect costs over the time range between project duration

limits is necessary for obtaining the minimum point of the combined curves.

-57-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
PHASE III - ALL-CRASH START

As demonstrated, the normal time-least cost point on the project

time-cost curve is one possible starting point for Phase III scheduling

variation. The other known point on the project time-cost curve at the

beginning of Phase III calculation is the crash time-all crash cost point

This may also be usee as a starting point. Referring to Figure 8, this

would mean starting at Point B rather than Point A.

jne_practical reason for starting at the all-crash point is that it

may be closer to the final scheduling solution than the normal point.

This could be the case where an owner agency has set a completion date

that only a nearly crash effort will satisfy. It is particularly advan-

tageous to start at the closest point to the final solution when calcula-

tions will only be made to the point satisfying the specified time

schedule. On the other hand, if the entire curve AC of Figure 8 is to

be developed, it is more advantageous to start at normal point A and

thereby eliminate the necessity for developing the curve portion BC.

A second and more important reason for developing a solution com-

menclng at the all-crash point is to present the mechanics for moving in

the opposite direction along the project time-cost curve. This procedure

involves changes in noncritical as well as critical operations. A detail

ed discussion of the entire solution is not necessary since it is similar

to that developed in the last section. Only those points that are

different or troublesome will be considered here.

The initial input data required from the estimator is the same as

that needed for Phase III - Normal Start. Therefore, Figure l0 may

again serve as the source of input data. The Summary Sheet, Figure l8,

has the same form as Figure ll. Two headings are varied slightly to

indicate that the maximum slope combination is to be used in place of

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

the minimum slope combination and that cycles may be limited by possible

lengthening of operations rather than possible shortening of operations.

On Figure l8, the eight initial cycles of Phase III involve lengthening

noncritical operations. This changes totaloroject costs but does_no£

-58-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
change project time. In addition there are the nine cycles previously

found necessary to develop the project time-cost curve when working in the

opposite direction.

The Operation Selection Sheet, Figure l9, has the same general form

as Figure l2. The principal difference is that the operations are arrang-

ed in order of decreasing cost slopes instead of increasing order. Many

of the numbers in column 3 are struck out during the lengthening process.

During the lengthening of critical operations in the later cycles, most

of the critical paths tend to become noncritical, which explains why the

numbers are struck out. On Figure l9 there are no "NC" entries in column

A as there were on Figure l2. A little thought will indicate that these

entries are not necessary in the lengthening process.

Figures 20A and B, the Interaction Limit Determination Sheets, are

the same as Figure l3 except that the word "lengthened" is substituted

for "shortened". The horizontal V symbol for indicating a change in

either an earliest finish date or an earliest start date is pointed toward

the right instead of the left. This is consistent with the visualization

of changes taking place in a corresponding schedule bar chart. As oper-

ations are lengthened, the necessary shifts occur toward the right, or in

the direction of longer time. As before, the absence of entries in a

column or entries in both halves of the column indicate no change in

the lag. A single entry in the^right half of the column indicates that,,

the earliest starting date of the following operation has shifted to a

later date, and the lag is increased. A^single entry in the left half of

the,_cn1umn, indicates, that the earliest finish date of the preceding opera-

tion has shifted tp,,a._la,ter, date,,atad the.J.ag^JLs decreased.

The Precedence Matrix, Figure 2l, is developed and used in the same

way as Figure l4. Only the initial form of the precedence matrix is
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

given. The final form at_the end of cycle l7 will be the same as the

initial form for t]iejs.horjtening^o^tion. Figure l4B. One part of the

procedure of updating the precedence matrix deserves comment. In the

lengtheningj>rocess it is more,.common for zero lags to change to positive

lags. This generally requires the removal of x entries as well as that

of the zero involved. Extreme care must be used to delete only the x's

-59-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
that were introduced solely by this zero. The first step is to remove

the appropriate x's in the row of the deleted zero. This requires back-

tracking downward one step to determine which x's were brought up by this

zero. If there are remaining zeros in the row of the deleted zero, then

a further check must be made to be sure that any x's being removed were

not also brought up as a result of these other zeros. If they were, they

should not be removed. Next, the x's in the related rows above that of

the deleted zero are removed in a reversal of the process by which they

were entered. Again, if there are zeros in such rows besides the one

related to the deleted zero, checks must be made to be sure that none

of the x's being removed are also present because of these zeros. If

they are, they should not be removed.

Figure 22 gives the initial form of the project network chart. The

final form would be the same as Figure l5A, the initial form for the short-

ening process. The lengthening process requires frequent remoyalj as_well

as addition, of lines in,this chart. When a zero lag becomes positive,

the appropriate line is removed from the chart. This may indicate that

a certain path ceases to be critical. If so, the lines that represent

the critical condition for this path must be removed. Moreover, the

corresponding entries in column 3 of Figure l9 must be struck out, and

the appropriate pair of parallel critical paths on the upper portion of

Figure 23 must also be struck out.

Figure 23 is the final worksheet for the lengthening procedure

and is very similar to Figure l6B for the shortening procedure. Non-

critical operations to be lengthened in the initial cycles are select-

ed one at a time. Therefore, this selection does__not_,require the use,

o£Jfrg*M««~^3 r Tn t-he first eight cycles^ the operations chosen were

^nJLcked_from those available in order of descending, cost slopes. This

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

is an arbitrary basis for the selection of noncritical operations, and

does not necessarily have to be followed. However, when critical op-

erations are lengthened, the procedure demands that a combination of

operations producing the greatest decrease in project cost per day

of project lengthening be chosen. To determine such a combination, the

highest available critical operation from Figure l9 is selected_asji,

first trial and an x entered in the appropriate space in the lower

-60-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
portion of Figure 23. By examining the parallel pair of critical paths

in the upper portion of the figure, the additional critical operations

that may be simultaneously lengthened are obtained. From each parallel

critical path the operation with the highest cost slope, that still can

be lengthened further, is selected and additional entries are made in the

column of the first x entry on the lower portion of Figure 23. On this

top portion of the figure, the circled operations are those having the

highest cost slopes rather than the lowest as on Figure l6B. When op-

erations on the upper portion have been lengthened to their limit they

J»re.-fi-truck- out. The fact that all operations in one path have been

struck out does not prohibit operations in the parallel path from being

lengthened. When a path heroines nonrrl tlcal^ the, pair of paths of which

it is one is struck out.

Having obtained the maximum possible number of operations, select-

ed as above, in a column on the lower portion of Figure 23, their com-

bined cost slope is totaled. Additional trials usually are necessary,

each starting with the next lower available and untried operation from

Figure l9. Trials cease when it becomes certain that no combination of

remaining operations can produce a higher cost slope than already obtain-

ed. One potential source of error should be pointed out. No operation

may be used in a combination when that operation appears in_jL,cr,itical

path thaj^ already contains one of the other operations of the combina-

tion.. For example, in the first trial of cycle l0 on the lower portion

of Figure 23, the combination of Operations 3, ll, and l4 gave the

maximum possible combined slope. Since Operation 3 is in a parallel

critical subpath with Operations 2 and l0, as entered in the top portion

of Figure 23 at the end of cycle 9, it would appear possible to also

lengthen one of these operations, say Operation 2, simultaneously. This,

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

in turn, would lead to the lengthening of additional operations, such

as Operations 4 and 6. However, both Operations 2 and l0 appear in a

critical path including Operation l4. This path has already been length-

ened as a result of lengthening Operation l4. Lengthening Operation 2 or

Operation l0 would not be the lengthening of a parallel operation but,

rather, of a series operation. This is not consistent with the procedure.

-6l-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Therefore these additional operations must be ruled out from the combin-

ation already selected. This possible source of error may be avoided

by the following procedural rule: When selecting,parallel critical path

operations that may be simultaneously lengthened, check to_be sure, that

they do not appear in paths containing operations already included in

the combination. If they do, they may not be used.

One further comment on the Phase III All-Crash Start solution should

be made. The results shown on the Summary Sheet, Figure l8, are incon-

sistent at one point with the results obtained by the Phase III - Normal

Start solution. Figure l8 indicates that the lowest project cost for

the project crash time of 49 days is $58,l90. Figure ll indicates that

the least cost for 49 days is $58,l30. The latter figure is the correct

figure. The inconsistency arises from the fact that there are alterna-

tive patterns for accomplishing the lengthening of noncritical operations.

The one obtained by the arbitrary rule of choosing operations in order

of decreasing cost slopes is not necessarily the best. It may result in

network relationships that permit certain noncritical operations to be

lengthened more and others less than an alternative pattern of selection.

After the first cycle of the lengthening of critical operations, a short-

ening cycle might be introduced. If the best possible pattern of length-

ening noncritical operations had not been used, then this shortening

cycle would result in a better network balance that would give a decreased

project cost at crash time. One or more pairs of a critical operation

lengthening cycle and a critical operation shortening cycle might be

necessary to establish the minimum crash cost, if it were desired. For

example, at the end of cycle 9 of Phase III lengthening, if a cycle of

* critical operation shortening were introduced, it would be found that

Operations ll, l4, and l5 would be shortened one day at a project cost
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

increase of $820. This would bring the time and cost respectively to

49 days and $58,l30. If critical operation lengthening were then resumed,

the exact nine cycles obtained in the Phase III - Normal Start solution

would be retraced in reverse. The procedure of alternately lengthening

and shortening critical operation combinations to obtain a more favorable

network balance for a given project duration is the basis for the method

of solution to be discussed in the next section.

-62-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
PHASE III - CONVENTIONAL ESTTMATE START-MECHANICS

At the beginning of Phase III calculations there are two determinable

points on the project time-cost curve. So far, procedures have been out-

lined to indicate how Phase III can be solved by starting at either of

these two points. A third procedure that starts at a random point not on

the project time-cost curve will now be discussed and its mechanics con-

sidered. Further attention will be given to its applications in the next

section.

Referring to Figure 8, it was previously pointed out that within the

area enclosed by curves ACBDA, there were an infinite number of possible

scheduling solutions for any project. Scheduling solutions that meet a

specified project duration time exactly are confined to points on a vert-

ical line through the area. Point E of Figure 8 is one scheduling solution

for a project time of 300 days. It is possible to move from point E to a

point on curve AC which gives the lea-Bt cost scheduling solution for a

300 day project duration and to recognize when this point has been reached.

Further movement in either time direction along the curve can then be made

by procedures already discussed.

Figure 24 gives an example of the data that would be required from the

estimator, plus some preliminary calculations. It is assumed here that the

estimator, using^conventional methods, has arrived_ajt a cost estimate and

aj^cpjTipajo^jjig_s_c^ They give a

direct cost of $57,350 for project completion in a specified time of 56

days. The operational times and costs that correspond to this estimate are

given in columns 2 and 3. These times and costs were selected to be con-

sistent with the data previously used in Phase III discussions. In other

words, the time and costs fall on the straight lines already used to rep-

resent operation cost variations between normal time and crash time limits.
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

The estimator also has been asked at the outset of Phase III calcu-

lations to consider each operation in the project and to develop two add-

itional sets of data. One set gives the shortest possible times for the

performance of all operations and the corresponding operation direct costs.

The second set gives the least costs at which operations can be performed

-63-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
and the corresponding operation times. It is assumed in the example that

the estimator arrives at the same data for "crash" performance and "normal"

performance as was previously used.

There are now three sets of estimated data representing, respectively,

a_Jfej^sJLble solution for performance in a specified .time, a crash solution,

and a least cost solution. The feasible solution probably does not lie

on the ideal project time-cost curve but, rather is equivalent to point

E of Figure 8. Its distance from the ideal curve will depend on the skill

of the estimator, as well as upon the validity of the assumptions on

which the ideal curve is based. From these three sets of data are esti-

mated cost slopes for both lengthening and shortening each operation.

Earliest start and finish times for each operation have been calculated

to provide information for later computation of lags. The earliest fin-

ish date of the final operation confirms the 56-day project duration.

Figure 25, the Summary Sheet, is similar to those previously used.

One column has been added to indicate, by initials, whether a cycle

involves lengthening or shortening an operation and whether it involves

critical or noncritical operations. The procedure is to first lengthen

all noncritical operations. This results in the equivalent of the vert-

ical line EF on Figure 8. The project cost is reduced without changing

the project duration. When the possibilities for lengthening noncritical

operations are exhausted, the next step is to leng_then that combination

of critical operations that produces the largest cost slojge. This is

illustrated by cycle 8 of Figure 25; the graphical counterpart on Figure

8 is line FG. Next, the project duration is Jbrought back to the specified

duration by shortening combinations of critical operations that produce

the leajst cost slope. This is illustrated by cycle 9 of Figure 25, and

the graphical counterpart on Figure 8 is line GH. This procedure of

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

"wiggling-in" is followed until the slope of a lengthening cycle is the

same as the slope of a preceding shortening cycle, or vice versa, The

point at which the slopes become the same marks the arrival on the ideal

curve.

In the example on Figure 25, all lengthening *nd shortening cycles

for critical operations were for one day at a time only. This was

-64-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
dictated by the limits of operational changes for this specific problem,

but need not be the case generally. On a critical lengthening cycle,

the,j>roject duration is changed by whatever limit is established by the

shorter of the operation lengthening limits or the interaction limit.

On the shortening cycle, the project duration is changed by the amount

necessary to bring it back to the specified duration. ThjLs_jnight require

more than one consecutive shortening cyclej_if one cycle failed to bring

project length back to the, specified duration. While it is suggested

here and illustrated on Figure 8 that the shortening cycles serve mere-

ly to bring the project duration time back to that specified, it would

also be perfectly permissible to carry a shortening cycle to its limit

at a less than specified project duration. In this case it may take

more than one consecutive lengthening cycle to return to the specified

time.

The Operation Selection Sheet. Figure 26, is slightly different than

those previously employed since it combines the functions of the corres-

ponding sheets from both shortening and lengthening solutions. The tally

portion of the sheet maintains a record of both the remaining possible

shortening and remaining possible lengthening for each operation. Column

A records separately when an operation has reached its limit of being

shortened and of being lengthened. Since the operations are arranged in

order of increasing cost slopes, selections will be made starting at the

bottom of the list on lengthening cycles.

The Interaction Limit Determination Sheet, Figure 27, is very simi-

lar to the sheets previously used. One additional row of headings has

been added to indicate whether the operations changed during a cycle are

being shortened or lengthened. Figure 28 presents the initial pre-

cedence matrix. The final precedence matrix would be the same as that
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

following cycle 5 of the shortening solution which is given on Figure

l4D. Figure 29 presents the initial project network relationships.

The final network relationships are the same as those given by Figure

15C.

Figure 30 is used to determine the appropriate combinations of

critical operations for both lengthening and shortening cycles. It is

a combination of the sheets used for the two previous solutions. The

-65-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
pairs of parallel critical paths are listed at both top and bottom of the

worksheet. Those at the top are used for lengthening cycles. The opera-

tions having the maximum cost slope in each column are circled. Operations

are struck out when they have been lengthened to their limit. At the

bottom of the sheet, circled operations represent those having the least

cost slopes in each column. Operations are struck out when they have been

shortened to their limits. The middle portion of the sheet contains the

trial runs for determining the optimum combination of operations. The

procedure is the same as previously explained, except that the bottom por-

tion of the sheet rather than the top is referred to during shortening

cycles.

-66-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
PHASE III - CONVENTIONAL ESTIMATE START - ADVANTAGES

The preceding three sections have presented the mechanics for a man-

ual solution of Phase III of the Critical Path Method beginning at any of

three possible starting points. This section suggests certain advantages

offered by the last procedure which begins with a schedule determined by

conventional planning methods. Since this is a new approach for the

Critical Path Method, it warrants a separate discussion.

The first advantage of the conventional estimate start is that CPM

is made a more practical tool at the estimating stage of a contract.

Slum l\m estimator is usually working under time pressure, his primary

goal is to determine feasible methods and costs early enough to enable

his company to submit a proposal by the bidding deadline. In any remain-

ing time, Phase III scheduling variations could be performed to improve

the estimate. This assumes that there is time to prepare the Phase I net-

work chart and time for the estimator to provide the additional time and

cost data required. Ideally, CPM might be used in making the original

estimate. A more logical stage of development, until CPM methods are

improved and contractors are better acquainted with them, is to use them

as secondary tools to improve the results of conventional estimates.

A second advantage of starting with a conventional estimate is that

the final solution may not be far removed from the initial solution. A

relatively few cycles of scheduling variation may give the ideal schedule

for the specified project completion time. This tends to make the non-

computer approach more practical. Moreover, as changes are made in pro-

ject scheduling, the project time either remains at the specified time

or varies only slightly from it. Since each change results in a cost

improvement, the process may be discontinued at any point and still give

a better solution for the desired project time.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

A third advantage of ,starting with ,a..cpnvent ion* 1 estimate is the

most important and most debatable? The estimates of data for shortening

or lengthening the various operations should be more realistic, and

therefore more accuratethan estimates made in the other two approaches.

This is based on the fact that the conventional estimate has established

-67-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
a realistic job pattern and setting. In general, methods and equipment

for meeting the required schedule have been chosen. The approximate

season of the year in which various operations are performed is known.

Decisions regarding overtime hours, length of work week, and shift work,

have been made. All of these are subject to change by application of

Phase III variations but it is unlikely that the overall pattern will

change completely. In this approach the estimator is asked to consider

each operation individually and to decide if it could be performed at a

lower cost if more time were available and whether it could be performed

faster if necessary. If so, he is asked to estimate both time and cost

for the least cost method and for the crash time method. Any single sched-

uling change would represent a relatively minor variation from the condi-

tion of the original feasible estimate. A series of such changes might

produce major differences in the job pattern but this can be handled ef-

fectively in a manual solution since the estimator has the opportunity

to revise his figures at the end of any cycle. Such a procedure involves

reasonable demands on the estimator and allows him to retain judgment

control.

Contrast this with the requirements of a computer solution beginning

at the project normal time-cost point. The estimator is first asked to

compute the least cost of every operation assuming that the entire project

would be performed at least direct cost. The fact that the resulting

"normal" project duration may be twice as long as the allowable project

duration means that the estimator is often figuring a completely different

job which has very little semblance to the one to be bid. Next the esti-

mator is asked to compute crash performance data for every operation. He

attempts to visualize an "all-crash" project and to determine the cost of

crashing each operation. Again he is dealing with an artificial job

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

situation that is generally quite a bit different than the realistic condi-

tions of the final solution. However, since all data is necessary as

computer input and since there is no opportunity to revise the figures

at intermediate steps of the procedure, the estimator must do the best

that he can. He is being asked to estimate data for two extreme situa-

tions without any "feel" for the conditions of the final scheduling

-68-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
solution and, further, to assume a straight line variation of costs over

the full time range between these extremes. It must be remembered that

this input data is the basis for the output information and th* resulting

decisions. The less realistic it is, the less valuable will be the de-

cisions .

It will be argued that cost-time variations for any operation can be

estimated without considering the performance mode of other operations

since CPM is based on the assumption that all operations are independent

of one another except for sequential relationships. Unfortunately this

is not true from a practical point of view. For example, crashing one

operation may require labor crews to work multiple shift, 6 days per week.

From a practical labor relations standpoint, crews on concurrent operations

can hardly be kept on a 40-hour week. It will also be argued that starting

at the end of the project time-cost curve is a more idealistic approach.

It therefore is conceivable that this approach will lead to completely

different results for a given project duration than are obtained by start-

ing with the conventional estimate data. Certainly the estimator will

be free from any bias if he does not consider the conditions surrounding

a particular, feasible solution. While this argument may have some valid-

ity, the conventional estimate starting point is recommended since it is

more likely to produce the most realistic final solution.

-69-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
EVALUATION AND APPLICATION

This report has presented noncomputer procedures for applying all

three phases of the Critical Path Method. In the introduction two just-

ifications were given for the use of such procedures. One was the incon-

venience of computer use for field organizations or small companies or,

sometimes, a lack of familiarity with their use. A second was that the

computer methods have certain shortcomings that can be overcome by a non-

computer approach. With a better understanding of the methods involved

in CPM, these justifications can now be explained and evaluated more

clearly.

Phase I of CPM is essentially a noncomputer procedure regardless of

the manner in which the other phases are handled. The network relationships

are developed, one by one, as they are represented graphically in chart

form. As new relationships are recognized, additions and alterations are

made on this chart. The arrow notation provides the dual numbering system

for each operation that is required by the computer to establish the oper-

ation sequencing relationships. The noncomputer approach depends on visual

observation of the network chart during various computational steps and,

therefore, can be satisfied by the single numbering system of the simple

circle notation. This circle notation makes Phase I easier by eliminating

dummy operations and permitting the graphical representation of operation

relationships in a more straightforward manner.

Phase II of CPM is a purely mechanical process once the network re-

lationships between operations have been established and the time of per-

formance of each operation has been estimated. Besides the network chart,

only a single array of data is required, and the calculations consist of

addition or subtraction of two numbers at a time. Even for a complex

project, the noncomputer procedure is practical to apply. Therefore, the

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

benefits of this phase are as available to the contractor who uses manual

methods as to the one who employs a computer.

Phase III of CPM requires a complex and lengthy set of calculations

even for a relatively simple example. It is for this phase that the com-

puter has its most obvious advantages. The procedures offered in this

-70-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
report permit a noncomputer solution of such problems. However, a strong

claim that the manual method is more convenient than the computer approach

can hardly be made in view of the effprt required to perform the manual

computations. The advantage of the manual method is that it overcomes

certain weaknesses of the computer solution which result from the assump-

tion that each operation in the project is independent of every other

operation except for the sequential relationships shown by the network

chart. Given the normal time and cost of each operation, the crash time

and cost of each operation, and the network relationships, the computer

proceeds to grind out the project time-cost curve data and the data nec-

essary to obtain corresponding schedules. This is exactly what was done

also in the noncomputer procedure illustrated in this report. However,

there is a basic difference. In the computer method it is impractical

to exert intermediate control over the calculations to recognize changes

in input data or related effects between operations. In the noncomputer

method, this intermediate stage judgment control need not be relinquished.

Changes may be made in the input data at the end of any cycle. Changes

may also be made that jointly affect several operations to reflect the

fact that changes in one operation often do affect other operations in

ways not reflected by sequential relationships alone.

To illustrate how changes in one operation affect another, consider

a project including the construction of a multi-span concrete trestle.

On a conventional bar chart, the scheduling of the construction of this

trestle might be shown as a continuous bar over a long duration of time.

However, at certain stages of trestle completion other operations may

begin. In this instance, CPM requires that this trestle work be broken

down into several operations such as: "Build First X Spans," "Build Y

Additional Spans," and "Build Last Z Spans." Suppose the Phase III anal-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

ysis indicates the advisability of speeding up the operation "Build F.rst

X Spans." One way of speeding up this operation would be to use a differ-

ent method or different equipment. However, from a practical standpoint,

the method and equipment cannot be changed for the first group of spans

without making the same change for the construction of the remaining spans

In this case, a change in one operation affects two other operations.

-7l-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Again, consider a slightly different plan. The operation "Build First X

Spane" might be speeded up by working a l0-hour day instead of using dif-

ferent methods or equipment. In this case, the later two operations are

not necessarily affected. However, there are other operations going on

at the same time as the building of the first X spans. From a practical

standpoint, it may not be possible to work a l0-hour shift on one opera-

tion without other crews working on concurrent operations also having a

l0-hour day. Therefore, this change in one operation, although it does

not affect subsequent operations does affect concurrent operations. A

final alternative might be to speed up the operation "Build First X Spans"

by simply adding more men to the crew for a temporary period of time. This

change has a still different effect since it doesn't cause variation in

any other operation. It is the only type of change that CPM theoretically

recognizes.

In most construction work, changes in one operation more often than

not affect other operations. Thus the cost slopes of many operations

tend to change as the overall project is compressed in time. If one op-

eration is shortened by adding an extra crane, say, to the job's equipment,

then that operation has been charged with the move-in cost for the crane.

Since the crane will be free part of the time and since it also can be

used on other operations to shorten thtm, the cost slopes of the other

operations will tend to decrease since they no longer have to pay the

crane's move-in cost. The noncomputer method allows these cost slopes

to be changed at this point. An alternative approach possible with the

noncomputer method is to consider the overall results of adding one crane

to the job. The time changes in all affected operations are computed

and one cycle of shortening is analyzed to determine the corresponding

time effect on the project. This project time change divided into the
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

cost of moving in the crane and paying job rental for the necessary per-

iod of time gives a combined cost slope for the decision to use an addi-

tional crane. This cost slope may be compared to other possible cost

slopes for shortening the job. This approach permits making the decision

to use the additional crane before the point is reached that the short-

ening of a single operation can absorb the total move-in cost and justify

-72-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
its addition. The computer methods for CPM do not satisfactorily handle

either of these approaches.

There are many decisions that the job planner roust make in changing

the project schedule that do affect several operations simultaneously in

ways not indicated by the network chart. The decision to work a nine-hour

day, to work a rflx-day week, to go to three-shift operations, to change

the originally proposed methods, or to bring in different types or sizes

of equipment are all examples of changes whose affects are not generally

confined to a single operation. The Critical Path Method offers the plan-

ner the mechanics for determining the effect of any of these decisions.

But it is only where the planner can proceed a step at a time and con-

tinually update his original data that he can most intelligently use these

procedures. Unfortunately the computer methods are not as flexible in

application. It is not practical to stop the computer at an intermediate

step and to reappraise the input data in terms of the scheduling changes

to that point.

Another planning opportunity is made possible by the noncomputer pro-

cedure and the opportunity to vary operation cost slopes. A good planner

will prepare manpower and equipment requirement charts to correspond with

his operation scheduling. It was mentioned earlier in the report that

the determination of float times in Phase II gave a basis for shifting

some operations schedulewise to take peaks and dips out of these manpower

and equipment charts. To the extent that some peaks and dips still exist

after noncritical operations have been shifted, further improvements are

possible. Often the decision is made to carry a number of men or to

retain pieces of equipment over a period when they are not required by

the schedule. The cost of this labor or equipment rental might either

be charged to overhead or to the operations that force this decision. In

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

either case there are, for a given period, men or equipment available at

zero cost. The operations in progress during this period should be

examined to determine if they could be shortened further, at zero cost,

by making use of this available manpower or equipment. Even ineffective

use, if it served to provide any shortening, would be profitable. As

a result, the schedule might be changed further to shorten project time,

-73-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
using zero cost slopes for certain operations. If project time had al-

ready been decreased to that desired, then certain other critical oper-

ations could be lengthened at a co6t reduction to obtain a more favorable

scheduling solution.

To sum up the evaluation of the noncomputer approach to the Critical

Path Method, it might be said that where the necessary computation can be

performed from start to finish in a purely mechanical fashion, the computer

approach is much more desirable. On the other hand, where the mechanics

of CPM can be coupled with the judgment of a skilled planner who is able

to continually update data and experiment with changes involving more

than one operation, then a superior project planning system is obtained

by the manual method.

The Critical Path Method is founded on three basic assumptions. Each

one is important, but, unfortunately, each one is seldom completely valid.

The first such assumption is that the project network is a realistic model

of project activity. It is very difficult to construct a project network

chart that represents every important sequencing relationship properly.

It is frequently easier to assume, for example, that one operation cannot

commence until another is completed, when actually they may overlap and

should be broken down differently. Unrealistic assumptions in construct-

ing the network chart affect all other CPM calculations. The second

assumption is that the linear time-cost relationship for each operation

is a satisfactory approximation. This report suggests two advantages of

the noncomputer procedure to improve this approximation. Using a feasible

conventional estimate as a starting point, a more realistic setting is

provided for making the initial approximation. Then, permitting adjust-

ments of the cost slope data at the end of any cycle allows the approxi-

mation to be further improved as the schedule takes shape. The third

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

assumption is that all operations in the network are independent of one

another except for network sequencing relationships. This is generally

not true. An advantage of the noncomputer approach is that it permits

proper consideration to be given to changes that do affect more than one

operation.

Practically all engineering methods depend on assumptions involving

-74-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
approximations. The effectiveness of these methods is measured by the

dependability with which results may be successfully applied. CPM is no

exception. If after a solution is obtained and final adjustments are

made, CPM gives consistently better results than those obtained by trial

and error or judgment alone, then the assumptions are justified.

-75-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
FUTURE IMPLICATIONS

The Critical Path Method is already an accepted tool of industry.

Its uses for planning, scheduling, and control have been indicated in this

report. In closing, a few predictions will be made concerning possible

future developments.

(l) The use of CPM in connection with equitable settlements of claims

for change orders and extra work orders has been mentioned. Not only does

CPM provide the mechanics for determining all of the operations affected

by a given change as well as the effect on project duration, but it provides

the mechanics for determining the least cost to compress project completion

time back to its original date if this is necessary. As jobs become more

complex and time limits more critical, as in the case of missile-base con-

struction, CPM will probably be used more extensively by contractors to

document claims. It is already being employed for just this purpose.* In

turn, it will become necessary for owner representatives to understand CPM

and, possibly, demand submission of project network charts as part of the

contract requirements in order to protect their interests. Understanding

CPM for this purpose implies more than mere acquaintance with the input data

and output data from computer programmed solutions. A basic knowledge of

the assumptions and approximations on which CPM is based will be necessary.

(2) An increasing use of CPM will tend to place more importance on

performance time data as well as performance cost data in contractors' cost

records. Conventional cost accounting provides unit costs for the various

operations that comprise the project. Usually there is little or no docu-

mentation to indicate relative speed of performance. On the other hand,

CPM recognizes that costs vary with the time allowed and that a crash op-

eration will have considerably higher costs than one performed at the nor-

mal production rate. A CPM schedule requires project operations to be

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

performed at various rates ranging from normal to crash. It is conceivable

that cost figures provided to management for control and to engineers for

future estimating may carry an accompanying indication as to whether the

"Missile-Base Builder Wins Profit With Paper," Engineering News-Record,

September 7, l96l.

-76-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
operation was performed at "normal production speed," "257, crash," or

"l007. crash" for example. But the solution is not this simple. The point

to be made here is that the intelligent use of job data for CPM planning,

estimating, or control will require some system of documentation to reflect

actual performance rates as well as actual costs.

(3) A rational approach to decisions regarding expenditures to lessen

risks is possible with CPM. For example, the probability that the river

level on a dam project will reach the cofferdam crest elevation at various

dates might be determined from hydrographic data. The cost of the result-

ing damage at these dates could be estimated. The probability multiplied

by the cost would indicate the "cost of the risk." CPM in turn provides

a method for computing the cost of speeding up the critical operations

to decrease this risk cost. A balance between the cost of further pro-

ject speed-up and the rate at which the cost of risk is being decreased

would provide another basis for scheduling. A similar approach might be

taken in avoiding liquidated damages charges on a project having a time

limit specified in calendar days. In the given number of calendar days,

the probabilities of having various numbers of working days may be esti-

mated. CPM could be used to determine the cost of shortening the job

in this range, and the rate of cost increase could be compared to the rate

at which the cost of the risk is decreased to obtain a logical scheduling

balance point. Such studies on a long duration project could be repeated

periodically as the probabilities involved were changed.

(4) From the contractor's viewpoint CPM provides a method for deter-

mining a schedule for meeting the owner's completion date at least cost.

In some cases the owner offers an incentive in the form of a bonus clause

if the contractor betters this completion time. In the absence of such

an incentive, there still may be some reasons to aim for earlier comple-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

tion. As mentioned earlier in the report, the contractor's ever increas-

ing overhead cost curve can be plotted against the project time-direct cost

curve obtained by CPM Phase III. A balance between the cost of speeding

up the project and the rate at which indirect costs are increasing may

indicate the desirability of earlier completion. The cost of lessening

certain risks may also influence the scheduling, as just discussed. Many

-77-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
intangibles may also be involved such as releasing key men for other work

or winning the respect of the owner by shortening the schedule. Present-

ly, however, there seldom seems to be sufficient incentive to make early

completion attractive to contractors. An executive of a chain of grocery

stores recently complained that completion dates of similar buildings

varied from 55 days to over 200 days. In such cases, there is certainly

a dollar value to the owner from earlier completion. Even for public

projects, early completion represents dollars savings. Many competent

contractors could provide such savings if given a slight incentive to do

so. The use of CPM in their job planning would be an additional tool for

this purpose. A bidding system that permits contractors to bid comple-

tion time as well as contract amount seems desirable. Some formula,

established in advance, for converting bids to a comparable basis and

for establishing liquidated damages would be necessary. The point is

that owners need a system for providing a realistic incentive for comple-

tion at that point in time that makes the total cost of the structure,

including the value of early completion, the least. With such incentives,

CPM can provide a more effective tool for cost reduction.

-78-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
BIBLIOGRAPHY

This bibliography contains references concerning the Critical Path

Method (deterministic case), PERT and PEP (probabilistic c.ise), and

related mathematical and scheduling topics.

l. Activity Definitions for Digital SIN'S MK III PERT Flow Chart, Marine

Division, Sperry Gyroscope Co., Division of Sperry Rand

Corporation, Syosset, N.Y.

2. Archibald, R.D., "The Role of PERT in WSPACS Working Paper AGC-2,

l5 October, l960.

3. Archibald, R.D., "The Utility of PERT in Production Evaluation and

Control", Aerojet General Corporation, l960.

4. Archibald, R.D. and Phelps, H.S., PERT Management Control System

Computer Program Write-up, Aerojet General Corp.,

Solid Rocket Plan, Sacramento, California, June l960.

5. Aviation Week, "AMC is Developing Program Control Plan", December l9,

l960, P.29.

6. Backer, Frederick, Jr., "A Discussion of Problems Involved in LESS

(Least Cost Estimating and Scheduling)" IBM Applied

Science, Dallas, Texas, l960.

7. Boyan, E.A., "Target Commitment Scheduling", Lecture Notes,

Course l5:7l, MIT, l946.

8. "Breakthrough", Special Projects Office, Department of the Navy-

v<\m on principles of PERT and its use on the Polaris

program. Merit Productions of California, l0044

Burnet Avenue, San Fernando, California, (28 minutes, $l50)

9. Breitenberger, Ernst, "Development Projects as Stochastic Processes",

Technical Memo K-33/59, December 1959, U.S. Naval

Weapons Laboratory, Dahlgren, Virginia.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

l0. Business Week, "Better Plans Come from Study of Anatomy of

Engineering Job", 2l March l959, pp. 60-66.

ll. Chipman, J.S. "PERT With Costs", Technical Report ll2 SRP, WSPACS

Working Paper No. 4, Aerojet General Corporation, l5 Feb. l961

79-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
l2. Clarke, Roderick W., "An Introduction to Critical Path Analysis",

Graduate School of Business, Stanford University,

March, l96l.

l3. Engineering News-Record, "Missile-Base Builder Wins Profit With Paper",

September 7, l96l.

l4. Engineering News-Record, "Perini Corporation Pioneers CPM; New Tool for

Job Management", Vol. l66, No. 4, January 26, l96l,

pp. 25-27.

l5. Fazar, W., "Program Evaluation and Review Technique", The American

Statistician, April l959.

l6. Fazar, W., "Program Evaluation and Review Technique", Statistical

Reporter, Bureau of the Budget, January, l959.

l7. Fazar, W., "Progress Reporting in the Special Projects Office",

Navy Management Review, April,l959.

l8. Flagle, CD., "Probability Based Tolerances in Forecasting and Plan-

ning", The Johns Hopkins University, Baltimore, Md., l954.

l9. Ford, L.R., Jr., and Fulkerson, D.R., "A Simple Algorithm for Finding

Maximal Network Flows and an Application to the Hitchcock

Problem", Canadian Journal of Mathematics, September, l957,

pp. 2l0-2l8.

20. Francis, H.G., and Pearlman, J., "PERT-Program Evaluation and Review

Technique", Internal Report, Light Military Electronics

Department, General Electric Corporation, l960.

2l. Freeman, R.J., "A Generalized Network Approach to Project Activity

Sequencing", IRE Transactions on Engineering Management,

September, l960.

22. Fulkerson, D.K., "Increasing the Capacity of a Network: The Parametric

Budget Problem", Management Science, Vol. 5, l959,

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

pp. 472-483.

23. Fulkerson, D.R., "An Out-of-Kilter Method for Minimal Cost Flow

Problems", Rand Paper P-l825, January l8, l960.

24. Fulkerson, D.R., "A Network Flow Computation for Project Cost Curves",

Management Science, Vol. 7, No. 2, January, l96l,

pp. l67-l78.

-80-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
25. Gass, S. and Saaty, R., "The Computational Algorithm for the Parametric

Objective Function", Naval Res. Log. Quart. 2, 39-46

(l955).

26. Applied Science Division, IBM, "Getting More for Less", Dallas, Texas,

l960.

27. Heavy Construction News, "Critical Path Tested at Little Long", Vol. 5,

No. 29, July, l960.

28. Muggins, William H., "Flow-Graph Representation of Systems", Operations

Research and Systems Engineering, Ed. by Flagle, Hugglns

and Roy, Johns Hopkin6 ?ress. Baltimore, l960.

29. Instruction Manual and Systems and Procedures for the Program Evaluation

System (PERT). Special Projects Office, Bureau of Naval

Weapons, Department of the Navy, Washington, D.C.

30. Jarrigan, M.P., "Automatic Machine Methods of Testing PERT Networks for

Consistency", Technical Memo K-24/60, August, l960,

U.S. Naval Weapons Laboratory, Dahlgren, Virginia.

3l. Kelley, J.E., Jr., "Computers and Operations Research in Roadbuilding",

Operations Research, Computers and Management Decisions,

Symposium Proceedings, Case Institute of Technology,

January 3l, February l, 2, l957.

32. Kelley, J.E., Jr., "The Construction Scheduling Problem (A Progress

Report)", UNIVAC Applications Research Center, Remington

Rand UNIVAC. Philadelphia, April 25, l957 (Ditto).

33. Kelley, J.E., Jr., "Extension of the Construction Scheduling Problem:

A Computational Algorithm", UNIVAC Applications Research

Center, Remington Rand UNIVAC, Philadelphia, November l8,

l958 (Ditto).

34. Kelley, J.E., Jr., "Parametric Programming and the Primal-Dual Algorithm",
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

Operations Research, Vol. 7, No. 3, l959, pp. 327-334.

35. Kelley, J.E., Jr., "Critical-Path Planning and Scheduling: Mathematical

Basis", Operations Research, Vol. 9, No. 3, l961,

pp. 296-320.

36. Kelley, J.E., Jr., and Walker, M.R., "Critical-Path Planning and Sched-

uling", Pxoc. East. Joint Computer Conference, l60-l73,

Boston, December l-3, 1959.

-8l-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
37. Kelley, J.E., Jr., and Walker, M.R., "The Critical Path Method of

Planning and Scheduling", Ideas for Management, Papers

and Case Histories presented at the l2th Annual Inter-

national Meeting of the Systems and Procedures Associ-

ation, Toronto, l960, pp. 403-4ll.

38. Klass, P.J., "PERT/PEP Management Tool Use Grows", Aviation Week,

November 28, l960.

39. Lasser, Daniel J., "Programmer's Guide to PERT, Phase I for the IBM

Type 709/7090. Missiles and Space Division, Lockheed

Aircraft Corporation, l960.

40. Malcolm, D.G., Roseboom, J.H., Clark, C.E., and Fazar, W., "Applica-

tion of a Technique for Research and Development

Program Evaluation", Opns. Res. 7, 646-669, l959.

4l Martino, R.L., "New Way to Analyze and Plan Operations and Projects

Will Save You Time and Cash", Oil/Gas World, September,

l959.

42. Martino, R.L., "How 'Critical-Path' Scheduling Works", Canadian

Chemical Processing, February, l960.

43. Mason, S.J., "Feedback Theory: Some Properties of Signal Flow-

Graphs", Proceedings of the IRE, Vol. 4l (September,

l953, ll44-56).

44. Mason, S.J., "Feedback Theory: Further Properties of Signal Flow-

Graphs", Proceedings of the IRE, Vol. 44, July, l956,

pp. 920-926.

45. Nieman, R.A., and Learn, R.N., "Mechanization of the PERT System on

NORC", Technical Memo K-l0/59, August, l959 (Rev. April

l960), U.S. Naval Weapons Laboratory, Dahlgren, Virginia.

46. Nieman, Ralph A. and Learn, Robert N., "Mechanization of PERT System
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

Provides Timely Information", Navy Management Review,

August, l960.

47. Oil and Gas Journal, "Critical Path Planning means More Economical

Turnarounds", June 6, l960, p. 78.

48. Ormsby, Snell, Tjersland, Vidalakls and Whitely, "Control of Product

Development Progress", Stanford University, Stanford,

Cal., l960.
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
49. Perlman, Jerry, Presentation on PERT System at General Electric

Light Military Electronics Division, General Electric

Light Military Electronics Division, Utica, New York,

July, I960.

50. PERT Summary Report Phase I, July, l958, Special Projects Office,

Bureau of Ordnance, Department of the Navy,

Washington, D.C.

5l. PERT Data Processing Handbook for Technicians, Special Projects Office,

Department of the Navy, Washington, D.C., June, l960.

52. Program Evaluation Procedure Instructions, Headquarters, Air Research

and Development Command, USAF; November, l960.

53. Program Planning and Control System, Special Projects Office, Bureau

of Naval Weapons, Department of the Navy, Washington,

D.C.

54. Progress in Resources Planning through PERT, Technical Information

Series R-60EML46, General Electric Light Military

Electronics Department, Utica, New York, June, l960.

55. Project Engineering, June l958, "Industry Borrows POLARIS Planning.

56. Project PERT, Phase II, Special Projects Office, Bureau of Naval

Weapons, Department of the Navy, Washington, D.C.

57. Reeves, Eric, "Critical Path Speeds Refinery Revamp, Canadian

Chemical Processing, October, l960.

58. Reynolds, A.J., "Here's the Course HEPC Took on CPM", Heavy Construction

News, Vol. 5, No. 29, July, l96l.

59. Salveson, M.E., "On a Quantitative Method in Production Scheduling",

Econometrica, October, l952.

60. Salveson, M.E., "The Assembly Line Balancing Problem", Journal of

Industrial Engineering, l955.

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

6l. Salveson, M.E., "Principles of Dynamic Weapon Systems Programming",

Willow Run Laboratories, The University of Michigan,l960.

62. Sayer, Kelley, and Walker, "Critical Path Scheduling", Factory,

July, l960.

63. Slnms, T.J.it., "The Critical-Path Method - a new approach to planning",

Engineering and Contract Record, June, l96l.

-83-
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
64. Stelnfleld, R.C., "Critical Path Saves Time and Money", Chemical

Engineering, November 28, l960. pp. l28-l52.

65. Vazsonyi, A., "The Role of Mathematics in Operations Research",

Ramo Woolridge Corporation, Los Angeles, l955.

66. Walker, M.R. and Sayer, J.S., "Project Planning and Scheduling",

Report 6959, E.I. du Pont de Nemours and Co.,

Wilmington, Delaware, March, l959.

-84-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
ACKNOWLEDGEMENT

This report was prepared under Contract No. NBy-17798 for the Bureau

of Yards and Docks, U. S. Navy. The title of this contract was "Applica-

tions of Operations Research and Other Cost Reduction Techniques to Con-

struction." Principal investigators for Stanford University were Professor

C.H. Oglesby and Associate Professor John Fondahl. This report is confined

to only one of several topics chosen for contract studies. The choice of

subjects for this contract was left entire1y to the principal investigators.

The Bureau of Yards and Docks, U. S. Navy, neither directed nor suggested

subjects for study or approaches to be taken.

The author is indebted to Professor Og1esby and Associate Professor

Richard Symons, Department of Civil Engineering, Stanford University, for

a careful review and many helpful suggestions during the preparation of

the first draft of this report. The author is also indebted to Mr. Pete

Lambert, formerly Chief Estimator for Pacific Bridge Company; Mr. Frank

Matthias, Vice President of Kaiser Engineers; and Mr. Tom Paul, Vice Pres-

ident of Peter Kiewit Sons Co.; for supplying information or application

opportunities that led to ideas that became part of this report.

-85-
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 OF
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

APPENDIXES

CHARTS AND TABLES

Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 PHASE I
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

APPENDIX A

PROJECT NETWORK
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
*&ifcpii* -riZirB** **sim!** (BvAfM

10 Gu^ua. Qtyiti*

rigure 1

Cast. Yard

and Shop

Manuf. Pre Cast

Caps

Job

Mobil.

Manuf. Pre Cast

.Deck Pane1s

Procure

Cone. Pi1es

®—

Procure

Stee1 Pi1es

d>—

Demo1ish

•^Exist.Pier

Exc. for Bui1d

Thrust Beam ...Thrust Beam

12)

„ Drive^

Cone Pi1esN

Procure Riprap

Drive

Fender Pi1es^
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

, Backfi11

Set Caps

-6f)

. Set

/Deck Pane1s

Insta11

Deck Hardware

(19)

CI.P Conc.v

for Deckand\

Caps
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Insta11

(|6) Bracing Girts

r , P1ace

V2) Riprap

Remove**

2\) Girts

Insta1r

Mech.

Services

C1ean

Up and|

Move

Out

Procure Rubber Bumpers

0 12 3

_5_

/Instal1

Rubber Bumper

) System

I0

II

I2

Sequence Steps

PROJECT NETWORK

rtpm 2
 rigor. •
3 Op7 5

10 II
Maur* }
Sequence Steps
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 PHASE II
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

APPENDIX B
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

CONSTRUCTION SCHEDULE
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
DETERMINATION OF CRITICAL OPERATIONS, TIME BOUNDARIES, AND FLOATS

Op

Est

Tim*

Stort

Finish

Fk

Mt

cm

Ear1iest 1 Latest

Ear1iest 1 Latest

Tota1

Free

Op.

15

15

30

30

20
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

20

12

17

II

7
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

24

17

41

II

19

II

15

10

10

15

25

II

II

41

52

12

25

34

13

14

30

44

14

21

25

46
DETERMINATION OF CRITICAL OPERATIONS, TIME BOUNDARIES, AND FLOATS

Op

tST.

3t

>9T

Free

Op

Time

Ear1iest

Lotest

Tota1

15

18

30

30

■J

20

28
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

17

19

II

13

41
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

43

19

43

24

15

35

20

10

25

28

II

52

54

12

34

44

10

13

44

44

</

14

46

49

3
 64
°TT

rttm >
JIIL
30 40 50 60

Working Days
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

CaIendar Dates J*
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

6* *i2 *26 7>> 7/\0 7/17 \A 7/5\ B,7 *U *2I *fee

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

APPENDIX C

PHASE III
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

SCHEDULING VARIATIONS - GENERAL

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 \
\

ritwM s
A* Normal

Project Time-Days
JIII1I
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

250 275 300 325 350 375

Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

PROJECT TIME-COST RELATIONSHIPS

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

APPENDIX D

PHASE III
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

"NORMAL START" PROCEDURE

 J
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
*

15

30

20

12

24

10

II

14

21

10

12

1500

7200

8400

2100

1400

800

6800

1000

600

3000
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

2500

1800

2600

8400

1900

1300

700

10

18

14
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

20

10

1500

8000

9000

2700

1560

1200

7800

1240

900

3450

3580

2700

3320

Difference |

JjjM Cost

TIME AND COST DATA-SLOPE DETERMINATION - PROJECT CRASH TIME

Normq1 Prod

Tin* Cost

15 |ro,8oq

10

5
OPERATION

SELI

"CTION AND TIME TALLY SHEET

Op

S1ope

$/Doy

CP

Cyc1e

Finish

Cyc1e

Norma1

Time

Poss

Short

Hevis<

Cyc1e 1

0 I im

e/m

maintn

q HJS

igrnr

%
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

'9

T 1—

12

II

40

1
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

50

30

28

10

27

24

22

20

18

12

4.

16

50

12

II

10

15
 .
/

o
o

5
4
3
2
9
7
5
4
1
3
2
9
7
5
4
3
2
1

20
19
17
15
14
13
12
10
20
19
17
15
14
13
12

6
I
X
O
O
O
O
8
O
O
6
O
O
<
•
O
O
I
8
6

16
16
18
16

II
II

Start
i<>
Post Op<;

Pl*ur* 14*
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

PRECEDENCfc MATRIX
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

PRECEDENCE MATRIX
 o
o
o
o
o
o

0
1
9
7
5
4
3
2
1
9
7
5
4
3
2
1

14
13
12
10
20
19
17
15
14
13
12
10

X
X
O
X
X
O
8
X
X
6
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
O
O
8
6

r>
18
16

II
II
End of CycIe I
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

PRECEDENCE MATRIX
 o
o
o
o
o
o
o

9
7
5
4
3
2
1
9
7
5
4
3
2
1

13
12
10
20
19
17
15
14
13
12
10

X
X
X
8
X
X
O
O
6
X
X
X
X
X
X
X
X
O
X
X
X
X
X
X
X
X
X
X
O
O
8
6

18
16

II
II
End of CycIe 6
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

PRECEDENE MATRIX
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 i
z

o
o
o
o
o
o
P
o
o
o
o
1
9
g
2

oz

01
71
91
91
01

6
8
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Z
X
X
X
X
X
X
X
X
X
X
X
X
6
8

<l

—.

j_
z\

61
81
p\
£1
II
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 18

Legend
Star!

rigm IS
Tlfm ISA

End of CycIe I
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

Connecting Line-Zero Lag =

Connecting Line-Critical Pnth=

Connecting Line - Positive Lag —

PROJECT NETWORK
PROJECT NETWORK
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 rigur* 1)D
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 rigur* lit
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
DETERMINATION OF MINIMUM

OPERATION SLOPES COMBINATION

Oper.

Sequence

Cycle 1

1-4

5-6

(D l

—®

7-l0

-7

1l-l4

15-l6

l7

l3 II

15 -

18

Op.

Slope

Cycle

40

XX

50

l6
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

50

l5

60

17

70

80

l00
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

l0

l50

160

13

l80

200

250

300

l2

300

II

360

14

400

8 x©|

Figure 16A
DETERMINATION OF MINIMUM OPERATION SLOPES COMBINATION

Operotton

Sequence

Cyc1« 1

Cyc1e?

Cyc1e 6

Cyc1e 7

Cyc1e 8

1-4

5- 6

7- 10

GO 1'

W '*>

t3 t4

K>:

(6>

i?

Note Fo11owinq Cyc1e 9, the

reroomtng operotion numbers

may be struck out See

report text exp1onation

II - 14

IS

IS- 16

fo) -

17

IB

Op
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

S1ope

Cyc1e

Cyc1e

Cy

Cyc1e

Cyc1e

Cyc1e

7
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Cyc1e

Cy|1e

40

XX

xx

XX

XXXX

XX

50

50

60

70

80

1 00

1 50

1 60

X XX

X XX

XX

16

XXX

X
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

APPENDIX E

PHASE III
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

"ALL-CRASH START" PROCEDURE

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
SUMMARY SHEET - PROJECT SCHEDULING ADJUSTMENTS

Cyc1e

Mo» Qp S1ope Comb

I4

I2

13

IS

#2

#4

£P_ Ross. Lengthening

2
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

♦2

» 3 |» 4

InterodI

'Limit

5
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

18

<N^C^^JChono^^|ecfcos1

Doys

400

300

300

200

I 80

I 00

80

60

Cost

1600

300

900

200|59

720

300 (58

240
NETWORK INTERACTION LIMIT DETERMINATION - SHEET I

Cyc1e

10

Operations

Lengthened

I4

12

13

15

1MJ5

3JT

14

jnte'oction^

Limit

Days

Lengthened"

18

T5sT

ES

■ re

Op

Op

Pre

EF

Log

Log
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

Lag

Log

Log

Log

Lag

Log

Lag

Lag

Log

8
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

no

II

12

13

14

IS

16

17

10

18

14

13

I3

33

I4

12

17

17

17

41

23

28

28

32

32
 o
o
o
o
o
o

9
7
5
4
3
2
1
7
5
4
3
2

12
10
20
19
17
13
15
14
13
12
10

—
—
—
—
X
6
X
8
—
O
—
6
X
X
X
—
O
X
X
X
X
X
X
X
X
X
X
O
O
8
6
I

18

—•
II
~x
~x
II
J*
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

Start @ A1I Crash

Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

PRECEDENCE MATRIX
DETERMINATION OF MAXIMUM OPERATION SLOPES COMBINATION

Operotion

Sequence

Cyc1e 1

Cyc1e 2

CycIe 6

Cyc1e 8

Cyc1e 9

1-4

5- 6

7-10

II- 14

15-.6

\L

©/

/Out-Cy.|^

18

/6ut-Cv.l6\

/6ut-cy.||\

^uf-Cvl^N

/Out-Cv.rK

Op.

SIope

Cyc1e 9

Cyc1e 10

II

12

13
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

14

15

16

17

14

11

400

360

300

300

250
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

200

1 80

160

150

100

80

70

60

XX XXX

X XX XX

XX XX"

XXXX

12

13

10

17

15

3
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

APPENDIX F

PHASE III
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

"CONVENTIONAL-ESTIMATE START" PROCEDURE

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
COST DATA-SLOPE DETERMINATION

Op.

Est

[Crosh

Crash'

Difference

Ca*t

S1ope to

Norma1

Difference

S1ope to

Eorhest

Eortesr

Time

Cost

Time

Com

T.me

Shorter

Tim*'

Cost

iT.me

Cost

Lengthen

Stort

Finish

5
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

1500

1500

none

1500

none

0
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

12

7680

10

8000

320

160

15

7200

480

160

12

25

8650

18

9000

350

50

30

8400

250

50

25
OPERATION SELECTION AND TIME TALLY SHEET

Op

S1ope

j/Doy

C. R

Cyc1e

Finish

Cyc1e

Est.

Tirm

Po«

Poes

Leng

Rev Time /

Remain Poss. Short. / Remain.Poss.

Lenc,

thening

15

Star

9
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

10

II

12

13

14

40

S 13

10
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

50

25

25

?5

24

16

50

S0

10

8
L3N

DM.

1 »d(

N0I1VNIWH3130. 1IWH N01lDVa31N

•• »|D*3

01

II

21

£1

tno|(oj»d0

fr1

£1

91

11'O!

ubi

SI
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

31'/

&

01

S'£

m p»BuD4T

01

>n

* iO B»U»IJO

1
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

p»n»i4|Dni

"UOI10DJ»|

oz

|IUI

buo

«*c

2
 z

o
o
o
o
o
o
o
o
o
o

0
9
5
2
1
9
5
1

02
91
51
21
01
02
91
51
01

W
O
X
X
O
O
6
8
X
X
Z
X
X
X
'
£
X
X
X
X
X
X
X
X
X
6
8
Z

z\

61
81
£1
61
81
£1

ZI
II
ZI
II
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

XIdlVW 30N3Q333bd
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

3|DUJI1S3 |DU01|USAU03 (D) |iDJS

DETERMINATION OF

OPERATION SLOPES COMBINATION

Op.

Sequence

Cyc1e 4

Cyc1e 7

Cyc1e 13

1-4

A Cg)

X®

5-6

5-

(§) ~

—-

0©

(2) —

<u

xz

7-10

ex

II - I4

d)

15- 16

- 16

_J
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

l7

@-

l6

Op.

S1ope

Cyc1e 9

Short.

Cyc1e II

Short

1213

5
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

I6

I5

I7

I0

I3

I2

II

I4

40

50

50

60

70

80

100

I50

I60

I80

200

250

300

300

360

400
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

APPENDIX G
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

OPTIMUM SOLUTION EXCEPTION

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
OPTIMUM SOLUTION EXCEPTION

A situation can arise where the manual procedures proposed by this report

for the Phase III scheduling variations will not produce the optimum solution.

Fortunately the probability that such a situation will occur is small and, if

it should occur, the variation from the best possible solution is generally

small. Therefore this shortcoming does not outweigh the advantages of pro-

viding a simplified approach that permits a manual solution to a complex

problem.

When scheduling variations are made according to procedures proposed in

this report, the operations changed in any one cycle are either all shortened

or all lengthened. Consideration is not given to the possibility of concur-

rently shortening some operations while lengthening others. However a simple

illustration will indicate a situation that could conceivably occur where

concurrent shortening and lengthening gives a better solution.

Figure 3l shows a portion of a project network chart. For the purpose of

illustration assume that these are the only operations in the network and that

their performance times and cost slopes are as stated on the figure. If the

project is being performed at least cost and normal time, the critical path

consists of Operations 5l-53-55 which require a project duration of 38 days.

To shorten this duration, Operation 53 would be shortened by two days at a

cost of $30 per day. At this point paths 5l-54 and 52-55 both become criti-

cal, and Operation 53 has been shortened to its crash limit.

Further project duration shortening using the procedure proposed by this

report would result in concurrently shortening both Operations 5l and 52 at a

total cost of $ll0 per day. However a better solution ic possible. Operations

5l and 55 can be concurrently shortened at a cost of .$l30 per day. Since both

of these operations lie on critical path 5l-53-55, that path is shortened two

days for every one day that the other paths are shortened. This would cause
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

it to become non-critical unless some operation on it that is not common to

the other paths is lengthened. Operation 53, previously shortened at an

expense of $30 per day, can be concurrently lengthened resulting in the recov-

ery of $30 per day. The net result of the combination of simultaneously short-

ening Operations 5l and 55 and lengthening Operation 53 is to shorten project

duration at a cost of $l00 per day. This is a better solution than that

giving $ll0 per day.

G-1
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
The existence of this situation may not be detected so easily in a com-

plex network. Work is underway to determine if an improved, but still simple,

procedure can be developed to give the optimum solution even when this special

combination of network relationships, operation times and costs does occur.

Note that the following conditions are required:

— (l) There must be at least three critical paths.

— (2) There must be some combination of operations common to

all critical paths and more than one of these operations

must be common to at least one of the critical paths.

(3) On a critical path that contains more than one of the

operations being changed there must be an operation that

is not common to any of the paths containing single opera-

tions being changed.

(4) It must be possible to change this operation, time-wise,

in the opposite direction.

(5) The net result of simultaneously varying some operations

in one direction and one or more in the opposite direc-

tion must be better than the result of varying the best

combination of operations in one direction only.

G-2
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 4

l5
20
55
70
20
54
30
10
53
10
52
50
51

8
8

80
28
60
16
ft/ Day

Crash Time
Operation

FIG. 31
Cost Slope

Normal Time
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

TIME AND COST DATA

Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

PORTION OF PROJECT NETWORK CHART

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

APPENDIX H
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

TO THE APPLICATION OF CPM

AN INFORMAL PRACTICAL APPROACH
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
AN INFORMAL PRACTICAL APPROACH TO THE APPLICATION OF CPM

Further work on the use of CPM since the first edition of this report has

strengthened the author's conviction that the application of the methods pro-

posed is both practical and sound. It is the purpose of this Appendix to elab-

orate on two points which were previously given insufficient emphasis.

The first concerns the advantages of the circle and connecting line

notation over arrow diagramming. The history of CPM indicates that early ap-

plications involved complex industrial plant problems having large numbers

of operations. Those responsible for the development included mathematicians

and operations research personnel. It is natural that a computer-oriented

system was developed. This includes arrow diagramming which is now univer-

sally employed by available computer programs. The proposal of a different

system for charting the project network should be based on a definite advan-

tage to the contractor. There is such an advantage. With a minimum of

instruction the man most familiar with the work may develop his own chart

rather than relaying information on sequential relationships to a skilled

specialist in CPM techniques who may understand little about the construc-

tion operations.

Consider the list of operations and sequential relationships shown on

Figure 32. The operation numbers represent tasks that the supervisory per-

sonnel understand and for which there are sequencing relations, as shown,

that they are best qualified to furnish. Now first attempt to draw a circle

and connecting line network and next an arrow diagram to chart a model of

this job. This problem was recently given to a group of contractors' person-

nel, many of whom were employing CPM in their work. In a couple of minutes,

all were able to draw a correct network using the circle and connecting line

notation. In a considerably longer time, none were able to develop the

correct arrow diagram. Solutions are shown on Figures 3 and 4 of this

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

report. The same problem was given to a group of 42 graduate engineering

students as an outside assignment permitting them to take as much time as

needed. Only four submitted a correct arrow diagram.

If a contractor's own personnel can easily draw network charts for their

projects they can also easily understand them and will accept them more

readily. If it then appears desirable to convert them to arrow diagrams in

H-l
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
order that existing computer programs may be utilized, then a specialist,

trained in these manipulations, can perform this task. The value of an

easily constructed network chart, lacking dummy arrows, and being simple

to revise, because arrow heads and tails do not have to be rigidly matched,

should not be underestimated.

The second point that will be discussed here is a less formal but more

practical approach to job planning and scheduling based on CPM techniques.

The procedures proposed in this report for the important and difficult

Phase III scheduling variations not only indicated the possibility of non-

computer methods but also indicated certain advantages of these methods

over the computer systems. However, the earlier discussion failed to stress

that in using these non-computer methods, good judgment may considerably

decrease the amount of effort required for sound planning and scheduling

decisions• A less formal approach to be described below is not as satisfac-

tory from a strictly theoretical mathematical standpoint because it does

not require a complete set of time and cost input data to be furnished at

the beginning of the solution process. Moreover it permits changes in the

data at any point in the process on a judgment basis. For these same

reasons that make this approach less satisfactory to the theorist, it be-

comes more useful to the competent planner and estimator.

The following suggested procedure utilizes a combination of CPM mechan-

ics and sound judgment to gain the most from both:

(l) Draw a network chart that is reasonably accurate. Avoid too

detailed an operation breakdown where it is recognized that

overlapping of operations may be possible. Indicate the

sequences that appear most realistic. A common problem here

is whether to base sequential relationships on physical

necessity or probable field scheduling restrictions. For

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

example assume that on a building job after the building

lines have been laid out, foundation excavation may be

started on all of three different wings immediately since

no other physical restrictions are present. However the

planner is reasonably certain that the same crew or equip-

ment will be Used on each wing, and he therefore antici-

pates excavating for Wing A first, then for Wing B, and

finally for Wing C. The question is whether the network

H-2
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
chart should be drawn based on this assumption or should it

indicate that all three excavation operations can simultan-

eously follow the building surveying operation? There are

two schools of thought on the proper choice. It is suggested

here that the original network chart be drawn to indicate the

excavation of the three wings as successive rather than con-

current, operations which is the most realistic sequencing

in the planner's mind at this time.

Having arrived at a network chart that appears to satisfac-

torily represent a realistic performance of the project, ob-

tain time estimates for each operation. These time estimates

need not be either the normal or crash estimates defined in

the report. Rather they should represent reasonable estimates

consistent with the time available for contract completion.

These would be the times used for bar chart plotting under

conventional planning procedures.

Perform Phase II calculations to determine the critical

operations, the floats of the non-critical operations, and

the resulting project duration. Since this is a purely

mechanical procedure, use an electronic computer if convenient.

If not convenient, careful manual calculation can furnish this

information within practical limitations of both time and

effort.

Knowing the critical operations, carefully examine them (and

also operations having very small total floats indicating

that they are nearly critical) to see if further breakdown is

desirable. Some operations may have been shown as end-to-end

tasks in the original network chart to avoid too detailed a

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

breakdown. Now that it is known that these operations are

critical ones, a closer examination is justified. By recog-

nizing that some of these operations may be overlapped and

should be broken down further, project time may be reduced

from that first calculated.

H-3
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
(5) To the extent that further breakdown of operations is desir-

able to indicate that some may begin before others are l007.

complete, add the resulting new operations and revise the

original network. Steps (l) to (5) 9hould be repeated as

necessary.

(6) Next, consider the possibility of decreasing project time

further without increasing project costs by reviewing Just

the critical operations. Generally these will only comprise

perhaps l5 or 20 per cent of the total operations. With the

knowledge that these are the operations that establish pro-

ject duration, the planner can afford to take a closer look

at them. Often it will be discovered that some of them can be

performed in shorter times than originally estimated without

changing their costs. There are several reasons that this is

so. First, having recognized their importance the planner

may devote more effort to arriving at faster ways to perform

these operations. Second, with the knowledge of the non-

critical operations in progress at the same time as any given

critical operation, the planner may consider the possibilities

of transferring men or equipment to expedite the critical

operation. While the non-critical operation might be extended

correspondingly, this would not affect project duration.

Third, some of the critical operations may be performed by

sub-contractors. In general, the sub-contractor has contrac-

tually agreed to adhere to the schedule established by the

general contractor. Within the range of reasonable demands,

he may be able to expedite his operations that are critical

ones.
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

(7) Having made alterations in the project network chart and

revisions to the time estimates of critical operations, Phase

II calculations are again repeated. The resulting project

duration may be either more or less than that specified by

the owner. Changes in critical operations should now be

made to bring the project duration in line with that speci-

fied. Here Phase III CPM mechanics can be profitably used.

However a complete set of time and cost data for all operations

H-4
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
is not necessary. Only those operations that are critical, or

that become critical because of the changes made, need be con-

sidered. Moreover, only the small portion of these that judg-

ment tells the planner may possibly accomplish his aim need

actually be considered. Therefore he generally will be dealing

with only a handful of operations rather than a very large

number. When critical operations must be shortened to reduce

project time, the sequencing decisions made in step (l) should

be reviewed. An economical means of shortening project duration

may be to change the scheduling restriction that seemed reason-

able at the outset. For example, at this time the planner may

decide that the foundations of Wings A, B, and C should be

excavated at the same time. The extra cost of additional labor

or equipment may be the most economical way to speed project

completion. This decision involves changes in the project

network.

Having made changes in the critical operations to bring the

project duration to that required, further changes should be

made to reduce project costs. This involves performance of

the procedure described in the report section on "Phase III -

Conventional Estimate Start." If necessary, Phase II calcu-

lations would be repeated once again to obtain new figures for

the floats of the non-critical operations. Then the planner

would carefully review the non-critical operations to see if

any might be performed more economically in view of the fact

that additional time is available. The fact that floats do

exist indicates possible dips in manpower or equipment

requirements if these non-critical operations are performed

Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

in the manner originally conceived. Recognizing that both

labor and equipment may have to be carried over these dips

for practical reasons, a review of methods is warranted.

Possibly a smaller crew, for example, can do this non-critical

job in the longer time permissible at a lower cost.

H-5
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
(9) Having exhausted possibilities of lowering costs by using more

time for non-critical operations, the "wiggling-in" procedure

for establishing a better balance between critical operations

can be employed. Again it should be emphasized that a complete

set of data is not required. The planner works with those crit-

ical operations that appear to give the results he is seeking

and utilizes CPM mechanics to determine when important network

relationship changes take place.

(l0) Once the planner is satisfied that he has reached the most

economical solution for the specified contract completion date,

he may proceed to investigate the merits of other completion

times. He may utilize CPM mechanics to develop a portion of

the project time-cost curve on either side of the specified

completion date. He may also develop an indirect cost curve

over this range. Again, this need not be a complete curve,

but need only indicate the incremental overhead costs associ-

ated with varying project completion over a relatively small

range. If applicable the planner would consider liquidated-

damage cost curves or bonus-penalty cost curves. The sum of

all of these curves over Just a portion of the time range

representing all possible project durations should indicate a

minimum total cost point. The corresponding project duration

sets a better goal for the proposed project schedule than

the completion date specified by the owner.

The foregoing procedure represents a practical approach to Job planning and

scheduling. It need not necessarily be carried through completely to produce

important improvements in the original schedule. At no point does it require

the estimator to furnish normal cost and time data and crash cost and time data
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970

for every operation as is required by the more formal computer solutions. It

does permit frequent review of the time and cost data that is being used to

determine if changes should be made. In summary, it takes advantage of the

skills of the good planner and estimator while at the same time furnishing him

a more powerful tool with which to work--CPM.

H-6
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Network Sketching Exercise

Sketch network model using:

(a) Circle and connecting line notation

(b) Arrow notation

Network relationships are as stated below:

Operation Must follow operation(s)

5l

6l

75

86

96

l0 2, 6

ll 7, 8

12 5, 9, l0

l3 ■ 3, l0

l4 4, l0

l5 l2, l3

l6 l3, l4

l7 ll, l4, l5

l8 l6, l7

Pig. 32
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
 67 354 A A 30
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated for 1 (Lafayette College) on 2014-03-14 12:52 GMT / http://hdl.handle.net/2027/mdp.39015000453970
Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google