Lectures 11

Process Monitoring and

Process Mining
Learning Objectives

• Understand how IT can be used to deliver value by automating business processes,

and how transactional data can be exploited to build insightful process analytics
• Apply qualitative and quantitative techniques to analyse process performance and to
assess the impact of process changes
• Apply the knowledge and skills acquired holistically on a real-life business problem.
Content

• Introduce common types of process performance dashboards, both for offline and
online process monitoring.
• Move to process mining techniques, which emphasize the use of process models
for process monitoring.
• Showing how these techniques provide a bridge from the monitoring phase back to
the discovery and analysis phases of the BPM lifecycle.
Exercise 11-12
Where are we?

Process
identification

Process
Process architecture
architecture

Conformance
Conformance and
and Process As-is
As-is process
process
performance
performance insights
insights discovery model
model

Process
monitoring and Process
controlling analysis

Executable
Executable Insights
Insights on
on
process
process weaknesses
weaknesses and
and
model
model their
their impact
impact

Process Process
implementation To-be
To-be process
process redesign
model
model

4
5

Business Process Monitoring

Performance Dashboards

Event stream
Enterprise
System

Process Mining
Database
Event log
Structure of a Business Process Event Log
What is a Process
Performance Dashboard?

7
Process performance dashboard
A visualization of a collection of process performance measures to put into evidence
information and patterns to help business users answer one or more business questions

Process
Dashboards

Operational Tactical Strategic

dashboards dashboards dashboards
(runtime) (historical) (historical)

8
Operational process dashboards

• Aimed at process workers & operational managers

• Emphasis on monitoring (detect-and-respond), e.g.:
- Work-in-progress
- Problematic cases – e.g. overdue/at-risk cases
- Resource load
 Tactical dashboards

• Aimed at process owners / managers

• Emphasis on analysis and management
• E.g. detecting bottlenecks

• Typical process performance indicators

• Cycle times
• Error rates
• Resource utilization
Tactical Performance Dashboard
@ Australian Insurer

11
Strategic dashboards

• Aimed at executives & managers

• Emphasis on linking process performance to strategic objectives

Manage Manage Manage Work Manage

Process Unplanned Emergencies & Programming & Procurement
Outages Disasters Resourcing
Key Performance
Customer 0.5 0.55 - 0.2
Satisfaction

Customer 0.6 - - 0.5

Complaint

Customer 0.4 - - 0.8

Feedback

Connection Less 0.3 0.6 0.7 -

Than Agreed Time

12
 Process: Manage Procurement 0.67

Process: Manage Emergencies & Disasters 0.58

Process: Manage Unplanned Outages

Overall Process Performance

0.54

1st Layer Customer Operational Risk Health

Financial People
Key Result Excellence Excellence Management & Safety
Area 0.5 0.4 0.65 0.5 0.8 0.4

2nd Layer Customer Customer

Key Performance Complaint Satisfaction
0.6 0.7

Customer Customer Average Time

Rating (%) Loyalty Index Spent on Plan
3rd & 4th Layer
Process Performance 0.7 0.6 0.8
Measures
Satisfied Market
Customer Index Share (%)
0.4 0.8
What does a performance dashboard consist of?

14
Process performance dashboards

A dashboard is a collection of Each widget displays at least one Four types of widgets:
widgets performance measure
E.g. The average case duration Single-dimensional widgets
Two-dimensional widgets
Three-dimensional widgets
N-dimensional widgets (tables)

15
Process performance dashboard widgets

• Single-dimensional widgets display

one variable

• We can group multiple measures in

one widget for convenience.

16
Process performance dashboard widgets
Two-dimensional widgets display two variables:

• Independent variable (x-axis) – variable with respect to which we want to analyze

• Dependent variable (y-axis) – the variable that we wish to analyze

Usually, a two-dimensional widget is a chart

• Bar chart, pie chart, etc.

Types of two-dimensional charts:

• Longitudinal charts: x-axis is time, y-axis is a performance measure

• Cross-sectional chart: x-axis is an attribute (e.g. activity, resource, country),
y-axis is a performance measure
• Histogram: x-axis is a performance measure, y-axis is a “count”

17
Process performance dashboard widgets
Longitudinal chart
Process performance dashboard widgets
Cross-sectional chart
Process performance dashboard widgets
Cross-sectional chart
Process performance dashboard widgets

• Three-dimensional widgets display three

variables, usually:
• Two cross-sectional variables
• A performance measure

• The third dimension can be coded in

different ways:
• Heat map: intensity and/or color
• Bubble chart: size and/or intensity
• Scatter plot: useful when the third dimension is
Boolean

21
Process performance dashboard widgets

• N-dimensional widgets display N

dimensions, usually:
• A cross-sectional variable
(e.g. the customer, the case identifier, etc.)
• Several performance measures

• In general, N-dimensional widgets

are tables

22
Designing Process Performance Dashboard
Identify a user or users and define a clear question or
set of questions that users will answer with this
dashboard

Identify the type of dashboard elements based on the

insights required to answer the question

Identify a type of visualization (e.g. type of chart) for

the element

Determine x-axis (independent variable): longitudinal

time, cross-sectional (attribute)

Determine y-axis (dependent variable): performance

measure or attribute and the aggregation function
Demo Time!
Tools for performance dashboarding

Operational-Tactical – Business Activity Monitoring (BAM)

• Axway AMPLIFY
• VITRIA Operational Process Intelligence
• Oracle BAM
• SAP Operational Process Intelligence (OPI)

Tactical
• Business Intelligence (BI) tools: PowerBI, Qlikview, Tableau…
• Process Mining (dashboard modules): Apromore, Celonis, QPR, …

Strategic Level – Balanced Scorecard tools

• BSC Designer Online, Quickscore, Sisense, etc.

25
28

Business Process Monitoring

Performance Dashboards

Event stream
Enterprise
System

Process Mining
Database
Event log
29
Process Mining
Automated
Process
Discovery discovered
process model

event log

29
Automated Process Discovery
Process Map
(directly follows graph)

CID Task Time Stamp …

13219 Enter Loan Application 2007-11-09 T 11:20:10 -

13219 Retrieve Applicant Data 2007-11-09 T 11:22:15 -

13220 Enter Loan Application 2007-11-09 T 11:22:40 -

13219 Compute Installments 2007-11-09 T 11:22:45 -

13219 Notify Eligibility 2007-11-09 T 11:23:00 -

13219 Approve Simple Application 2007-11-09 T 11:24:30 -

13220 Compute Installements 2007-11-09 T 11:24:35 - BPMN process model

… … … …
Process Map (aka Directly Follows Graph)

A process map of an event log is a graph where:

• Each activity is represented by one node
• An arc from activity A to activity B means that B is directly followed by A in at
least one trace in the log
 Anatomy of a Process map
Activity self-
Eventually-follows relation loop
between two activities (rework) Final activity
(not visualized) (multiple
possible)
Activity
Initial activity
(and its
(multiple possible)
frequency)

Short loop
(possible
rework)
Directly-follows
relation
between two
activities
(and its frequency)
Process maps are easy to understand, but they can be misleading!
32
Example: Process Map
with frequency

a
Event log
5
5: a,b,c,g,e,h
5: a,b,c,g,e,h b
4: a,b,c,f,g,h 5
c
3: a,b,d,g,e,h
5
3: a,b,d,e,g,h g
5
e
5
h
Example: Process Map
with frequency

a
Event log
5+ 4
5: a,b,c,g,e,h b
4: a,b,c,f,g,h 5+ 4

4 c
3: a,b,d,g,e,h
5
4
3: a,b,d,e,g,h f g
5
4 e
5
h
Example: Process Map
with frequency

a
Event log
5+ 4+ 3
5: a,b,c,g,e,h b
4: a,b,c,f,g,h 5+ 4 3

4 c d
3: a,b,d,g,e,h 3
5
4
3: a,b,d,e,g,h f g
5+ 3
4 e
5+ 3
h
Example: Process Map
with frequency

a
Event log
5 + 4 + 3+ 3
5: a,b,c,g,e,h b
4: a,b,c,f,g,h 5+ 4 3+ 3

4 c d
3: a,b,d,g,e,h 3
5
4
3: a,b,d,e,g,h f g 3
3 5+ 3
4+ 3 e
5+ 3
h
Example: Process Map
with frequency

a
Event log
15
5: a,b,c,g,e,h b
4: a,b,c,f,g,h 9 6

4 c d
3: a,b,d,g,e,h 3
5
4
3: a,b,d,e,g,h f g 3
3 8
7 e
8
h
Process Maps
• Commercial process mining tools use process maps as the main visualization
technique for event logs
• These tools also provide three types of operations:
1. Abstract the process map:
2. Filter the traces in the event log…
Examples

Patient Treatment Process @ Hospital (Sepsis IT Incident Management @ Bank

infections)
Process map: Abstraction
Common process mining tools provide abstraction capabilities on top of
process maps, to:
• Show only most frequent activities
• Show only most frequent arcs

Some tools offer further abstraction capabilities to:

• Show only least frequent activities
• Show only least frequent arcs
• Show only fastest/slowest activities
• Show only fastest/slowest arcs
Log filters

• Case-Level Log Filters

• Retain (or remove) all cases that fulfill a given condition, for example all cases that take more than 3 days

• Event-Level Filter
• Retain or remove individual events (activity instances) that fulfill a given condition, e.g. all activity instances
that were performed by “John Smith”

• A case-level filter either retains an entire case or it removes the whole case while an
event-level filter may remove some events in a case, and remove others.
Case-Level log filters

Attribute filters
• Retain only cases such that at least one event fulfils a condition over one or more attributes, for example: at
least one event where Activity = “Close order” and Resource = “Anti Alman”

Timeframe filter
• Retain traces that start, end, or are contained in a time period (e.g. all cases that started in January)

Performance filter
• Retain only cases that have a duration above or below a given value (e.g. 3 days)

Path filter
• Retain traces where there is a pair of events that fulfil a given condition (e.g. “Create invoice” eventually
followed by “Create purchase order”)

Repetition filter
• Retain traces that start with or finish with an event that fulfils a given condition
Event-Level log filters

Attribute filters
• Retain only events that fulfil a condition, for example: at least one event where
Activity = “Close order”

Timeframe filter
• Retain all events that happen during a certain time period (e.g. all events in
January)

Between filter
• For each case, retain all events between the (first or last) occurrence of an activity
and the (first or last occurrence) of another activity (e.g. between “Create PO”
and “Approve PO”)
Demo Time!
 Limitations of Process Maps

• Process maps over-generalize: some paths of a process map might not exist and
might not make sense
• Example: Draw the process map of [ abc, adc, afce, afec ] and check which
traces it can recognize, which don’t make sense given the log.
• Process maps make it difficult to distinguish conditional branching,
parallelism, and loops.
• See previous example… or a simpler one: [abcd, acbd]

• Solution: automated BPMN process discovery

Process Mining Using BPMN
• In addition to discovering a process map from an event log, process mining tools
allows us to discover BPMN models.
• This allows us to distinguish between conditional branching, parallelism, and loops
• To get an intuition of how this works, let’s draw the process maps of the following
logs and let’s try to imagine what is the BPMN process model corresponding to these
logs:
• [abc, abd]
• [abcd, acbd]
• [abcd, abcbcd]
1. [abc, abd]
2. [abcd, acbd]
3. [abcd, abcbcd]
 D
Discovery of BPMN models

A given configuration of node/arc sliders, and parallelism slider

Note: Apromore uses an algorithm called Split Miner to turn process maps 52
Demo Time!
54
Process Mining
Automated
Process
Discovery discovered
process model

event log

Conformance
Checking /

process model or compliance rules

54
Conformance Checking (model-based)

Event log:
ABCDEH
ACBDEH
ABCDFH
ACBDFH
ABDEH
ABDFH

Unfitting behaviour:
• Task C is optional (i.e. may be skipped) in the log
Additional behavior:
• The cycle including IGDF is not observed in the log
Conformance checking (rule-based)

• Oftentimes, we don’t have a full process model available

• Instead, we have a set of compliance rules, e.g.
• Certain tasks MUST be executed in every case
• Certain tasks can be executed at most once
• When a task A occurs, another task B must also occur (within e.g. 2 hours)
• Two tasks A and B CANNOT be executed by the same person (four-eyes principle)
Demo Time!
59
Process Mining
Automated
Process
Discovery discovered
process model

Performance
Mining
event log
Enhanced
process model

Conformance
Checking /

input process
59
Performance Mining
• Process mining tools typically support performance mining in three ways:
• Performance-enhanced process maps
• Dashboards
• Log animation

60
Peformance-Enhanced Process Map

Activity label
(and its duration)

Directly-follows relation between two

labels (with waiting time)

Bottleneck node
(all incoming
arcs are slow)

61
Demo Time!
64
Process Mining
Automated
Process
Discovery discovered
process model

Performance Variants
Mining Analysis
event log
Enhanced Difference
process model diagnostics
event log’
Conformance
Checking /

input process
64
Variant Analysis
Fast cases Slow cases
Find differences in 75% 25%

performance and control-

flow between multiple
variants of a process

A process variant is a subset

of the executions of a
process corresponding to a
given product, customer,
etc.
Case Study: Suncorp Group

• General & life insurance, banking, superannuation and

investments management
• 9M customers
• 16K employees
• $85 billion in assets
Suncorp Insurance

End to end insurance process

Produc 500
Sales Service Claims
t Dev tasks
Source: Guidewire reference models

Each process is varied by product & brand…

Home       30
Motor
Commercial


 



  



variations
Liability     
CTP / WC     

Total process variants: 3,000+

Case Study: Variants Analysis at Suncorp
Expected
Ba Performanc
d e Line

OK

OK Good
 Variants Analysis via Process Map
Comparison

Simple claims and quick Simple claims and slow

S. Suriadi et al.: Understanding Process Behaviours in a Large Insurance Company in Australia: A Case Study. CAiSE 2013
Variant Analysis

In process mining tools, such as Apromore, variant analysis can be

approached as follows:
1. First, use filtering to “slice” the event log into multiple logs (one per
variant).
2. Then compare the multiple logs using:
 Side-by-side comparison of process maps or process models
 A comparative multi-log dashboard
 A multi-log animation
Demo Time!
http://apromore.cs.ut.ee
Summary: Four Process Mining Capabilities
The bulk of process mining techniques can be divided into four capabilities.

EVENT LOGS

Automated Process Conformance Variant Analysis Performance Mining

Discovery Checking

By using the automated Conformance checking Process outcomes often By applying performance
process discovery, an provides a comparison of differ in practice. But mining techniques, e.g.
actual process model is the actual and the target why? nodes or cycle times can
created from the event process and thus The variant analysis be highlighted in color to
log. This represents the enables the detection of enables a simple simplify the subsequent
current sequence of differences and comparison of several analysis.
work steps and does not deviations. process variants and
necessarily have to serves as the basis for
correspond to the target process optimization.
process.

75