Introduction

- extremely fast cycle time

- focus on what customers want (without asking them)
- scientific approach to making decisions

five principles
- entrepreneurs are everywhere (any company size, sector or industry)
- entrepreneurship is management. a startup is an institution, not just a product.
- validated learning/experiments.
- build-measure-learn. turn ideas into products, measure how customers respond, and
learn whether to pivot or persevere. this feedback loop needs to be as fast as possible
for startups to succeed.
- innovation accounting. how to measure progress/set milestones/prioritize work.

Part 1 - Vision
Startups require management.

Lean startup is adapting the principles of toyota lean manufacturing to the highly uncertain
context of entrepreneurship.

The goal of the startup is to build the thing customers will pay for as quickly as possible. so
milestones should be measured by learning and not by physical (or digital) products built.
constant adjustments instead of rigorously executed
plan based on a lot of assumptions.

the product changes often through optimisation. the

strategy may change sometimes sharply, in what is
known as a pivot. the vision should remain constant
throughout.

look into books on how to develop a vision

a startup is a portfolio of activities: acquiring new

customers/serving existing ones, tuning product,
marketing, operations etc. a tiny startup faces the
challenge of serving existing customers while trying to
innovate. a massive company must keep investing in innovation lest it becomes obsolete. as
companies grow, what changes is the mix of these activities in the company’s portfolio.
a failure to deliver results is either a failure to plan adequately or execute properly. both are
significant lapses, but necessary in modern product development. internal
innovators/intrapreneurs are entrepreneurs as well, and can benefit from entrepreneurial
management.

Define
an entrepreneur is anyone with a clear vision of the future and a desire to innovate, roughly
speaking.

A startup is a human institution designed to create a new product or service under conditions of
extreme uncertainty. ← most important part of this definition is what it omits (industry sector,
size of company, etc)

A startup is greater than the sum of its parts, it is a human enterprise. it involves hiring creative
people, coordinating their activities, and creating a company culture that delivers results.

read the Innovator’s Dilemma at some point (Intuit’s SnapTax story). SnapTax succeeded due to
a process deliberately facilitated by senior management. It started with the idea of tax return
data collected automatically, and initially launched a mobile phone product (after speaking to
customers) that worked only in California and only on a very specific tax return. But big success.
Intuit then went from one initiative every tax season to 500 tests across a two and a half month
long tax seas, adopting the Lean Startup Way.

Leadership requires creating conditions that enable employees to do the kinds of

experimentation that entrepreneurship requires.

Learn
Lean thinking defines value as providing benefit to the customer;anything else is waste. In a
manufacturing business, customers don’t care how the product is assembled, only that it works
correctly. But in a startup, who the customer is and what the customer might find valuable are
unknown, part of the very uncertainty that is an essential part of the definition of a startup.

IMVU experimentation was really based on analysing customer use of website and product,
occasional focus groups. A/B testing etc. [And you can’t really do that testing if you don’t have a
product, even if it’s not very good.]

The irony is that it is often easier to raise money or acquire other resources when you have zero
revenue, zero customers, and zero traction than when you have a small amount. Zero invites
imagination, but small numbers invite questions about whether large numbers will ever
materialize. But validated learning can counteract the ‘audacity of zero’ - in the end, even if you
spend a lot of money fuelling ‘growth’ ie image, the fundamentals of your business will
eventually catch up and burn you out.

It is important to stress at this point that lean startup does not mean launching a low quality
prototype, charging customers from day 1 and using low volume revenue targets to drive
accountability. this may not work for some industries, sectors, applications or customers (ie you
would not give the US military an application full of bugs for ‘testing purposes’). In other words,
Lean Startup is a methodology not to see if the product can be built - all products can be built in
the modern economy - but if the product should be built, and if a sustainable business can be
created around that particular set of products and services. To answer these questions, a
method for breaking down a business plan into its component parts and testing each empirically
is required. so in Lean Startup, everything a company does (product, feature, marketing
campaign) is understood to be an experiment designed to achieve validated learning.

Experiment
if you cannot fail, you cannot learn. just shipping a product without a plan will not necessarily
mean you gain validated learning. a startup experiment should follow the scientific method - it
begins with a clear hypothesis of what should happen, based off of the startup’s vision (in the
same way a scientific experiment’s hypotheses are driven by theory), and then tests those
predictions empirically.

Zappos, for example, began with the vision of a superior online retail experience for purchasing
shoes. The founder took pictures of local stores inventories, posted the pictures online, and
promised to come back and buy them at full price if a customer placed an order. Doing so he
tested if there was existing demand for an online shoe shopping experience, but also tested
other things ie interacting with customers, payments etc. this is different from market research -
a survey would have informed him what customers thought they wanted, but his simple product
told him what real customers really were after.

Break it Down
the first step is to break down the grand vision into its component parts. the two most important
assumptions entrepreneurs make are the value hypothesis and the growth hypothesis.

● the value hypothesis tests whether a product or service really delivers value to
customers once they are using it. experiments offer a better indicator than surveys,
because people have a hard time assessing what they want objectively. for example, if
you wanted to get more employees within your multinational to volunteer, you could start
with a small number of volunteers and then see their retention rates. if they sign up again
on their own, that is a strong indicator they would find a volunteering program valuable.
● the growth hypothesis tests how new customers will discover a product or service and
can be tested in a similar way. for example if testing for viral growth, identifying early
 adopters and observing their behaviour (if they bring any friends onboard) would be one
way to see if the product or service would grow virally. again back to the volunteer
example, if all early adopters decline to sign up again (even after you’ve ensured they
get the best possible experience aligned with their values) then it’s time for qualitative
feedback and figuring out where to improve. as an advantage over traditional market
research, running this experiment we already have a cohort of people we can talk to to
improve and have already measured their behaviour. also, these experiments can be run
in parallel to the strategy actually being formulated. failures in one experiment can alter
the assumptions underlying your strategy.

kodak gallery head of product uses these 4 questions to make the division more lean:
1. Do consumers recognize that they have the problem you are trying to solve?
2. If there was a solution, would they buy it?
3. Would they buy it from us?
4. Can we build a solution for that problem?”

also, importantly, they build a product to test these hypotheses and then improve it based off of
the feedback they gather.
in relation to how Obama’s consumer federal protection bureau was set up: treat the CFPB as
an experiment, identify the elements of the plan that are assumptions rather than facts, and
figure out ways to test them. Using these insights, we could build a minimum viable product and
have the agency up and running—on a micro scale— long before the official plan was set in
motion.

Part 2 - Steer

how vision leads to steering

a startup is a catalyst that turns ideas into products. their products are experiments. interactions
with customers give feedback, both qualitiative (like/don’t like) and quantitative (how many
people). learning how to build a sustainable business is one of those experiments.

the two most important leap of faith assumptions (ideas)

are the value proposition and the growth proposition.
these give rise to tuning variables that control the
startup’s engine of growth. each iteration of a startup is
an attempt to rev this engine to see if it will turn.

once clear on these assumptions, you must enter the

build phase as quickly as possible with a MVP. the MVP
is the version of the product that enables a full turn of the
build-measure-learn loop with a minimum amount of
effort and the least amount of development time. it may lack many essential features the full
product will contain later, but it can also be more work as you need to be able to measure its
impact. you need to get the MVP in front of potential customers to gauge its impact.

when you enter the measure phase, the biggest challenge is determining whether product
development is leading to real progress. learning milestones are useful for entrepreneurs to
measure progress, and to investors for holding them accountable.

once the loop is completed, the entrepreneur faces one of the most difficult questions they can
face: to pivot, or not to pivot. if we’ve discovered that one of our hypotheses is false, it’s time to
make a major change.

Leap
for startups, the role of strategy is to figure out the right questions to ask. something about
analogs and antilogs (steve jobs knew people were willing to listen to music in public because of
the sony walkman - sony did not know that. steve jobs knew that people were willing to
download music, but not pay for it - napster was the antilog).

for every entrepreneur in the right place at the right time, there are hundreds of others also there
who failed. even facebook had college social networking competitors. successful entrepreneurs
had the foresight, ability and tools to discover which part of their plan was working and which
parts needed to be adjusted. understanding what’s driving your growth and your value is
important, as there are a lot of different potential reasons, and some are destructive - ie
investors pouring in money constantly and lots of paid advertising, but no value is actually being
generated. so traditional accounting has its limits - innovation accounting (to be discussed in
detail later) is necessary.

to do so, you must ‘go and see for yourself’ (genchi gembutsu - from the toyota way, lean
manufacturing - for the 2004 sienna minivan, toyota’s chief engineer went on a road trip across
all of america, mexico and canada, talking to users along the way). all successful sales models
rely on breaking down monolithic organisations into the disparate individuals that make them up.

design and the customer archetype

contact customers early, not with a product solution, but to clarify at a basic level that we
understand our potential customer and the problems they have. With that understanding, we
can craft a customer archetype, a brief document that seeks to humanize the proposed target
customer. This archetype is an essential guide for product development and ensures that the
daily prioritization decisions that every product team must make are aligned with the customer to
whom the company aims to appeal. Lean UX is a variation of UX design that assumes that the
customer archetype is provisional until validated learning has proven the business can serve
this type of customer sustainably (look into this).
there are two risks here: just-do-it, and analysis paralysis. the right balance between these two
things is called MVP, and is discussed in a subsequent chapter.

Test
MVPs aren’t necessarily the smallest product imaginable, but the product that will get you
through the entire build-measure-learn cycle the fastest. early adopters are key here, as they
much prefer an 80% product. they’re suspicious of things that are too polished, because they
want to be the first ones to use something, for the sake of being first (or possibly gaining
competitive advantage, in enterprise sales). to judge how robust the MVP needs to be - when in
doubt, simplify. most people dramatically overestimate features.

dropbox’s MVP was a 3 minute video of him explaining what he’d like to do. the website’s beta
testing list went up from 5k people to 75k people overnight.

food on the table went out and did market research, and at the end of the focus session they
pitched a sale - they got rejected by all but one customer. and what they did with that customer
was concierge service - the CEO went to her house personally to deliver recipes and
ingredients, as opposed to the fancy website plan they had in their mind. and each week, he got
feedback and improved on his product - with each new customer that got concierge service,
they got to know the local grocery store better. their product improvements too were marginal:
before a website, delivering recipes by email. then parsing ingredients list through software
rather than by hand. then taking credit card payments. the only way to know if your business
model is viable or not is to test its growth systematically with real customers.

the guys behind aardvark built six different products and tested each of them with real beta
customers. they used humans to replicate the back-end as much as possible. each product took
them two to four weeks to churn out. once they settled on a product, they brought in six to
twelve people a week to use their product. their engineers would attend these sessions. and for
more challenging technical problems, they used wizard of oz testing - people thought they were
interacting with the product, but in reality behind the scenes real humans were doing the work.
that way they were testing that if they overcame the technical problems around using AI to
answer questions naturally, people would use it.

the role of quality and design in a MVP

if we do not know who the customer is, we do not know what quality is.

if a MVP is perceived as low quality by customers, we can use this as an opportunity to learn
what attributes the customer cares about. this is much better than whiteboard strategizing
(speculation).
customers don’t care about how much time something takes to build. they only care if it serves
their needs. IMVU first shipped a MVP in which 3d avatars were stationary, as making them
move fluidly cost a lot of money and time - customer feedback consistently specified they
wanted the avatars to move. so instead of developing fluid movement, they shipped a MVP in
which the avatars teleported, without any fancy animation - a cheap hack. but customers were
super happy with it.

As you consider building your own minimum viable product, let this simple rule suffice: remove
any feature, process, or effort that does not contribute directly to the learning you seek.

the most common issues in building a MVP are: legal issues, fears about competitors, branding
risks, morale impact.
● legal issues can arise in patent-protected industries, such as a startup that relies on a
major scientific breakthrough. seek legal counsel before building a MVP.
● competitors will not steal your idea, especially established companies. MDs are flooded
with good ideas on a daily basis - their role is prioritization. it’s so hard to get noticed as
a startup, let alone ‘stolen’. sooner or later startups will face competition from fast
followers - you are doing a startup because you believe you can move through the
build-measure-learn feedback loop faster than anyone else. if your startup fails once one
competitor exists, you’re doomed anyway.
● don’t worry about branding - as a startup you don’t have any press and very few
customers. use this as an opportunity to experiment under the radar. engage in pr once
you have a product which you have tested with real customers which you are convinced
works.
● entrepreneurs must possess a mix of perseverance and flexibility - you need to commit
to a cycle of iteration before beginning development. MVPs often bring bad news, as
they are a (much needed) dose of reality. when that happens, it is time to pivot. but not
to call it quits.

Measure - Innovation Accounting

A startup’s job is to (1) rigorously measure where it is right now, confronting the hard truths that
assessment reveals, and then (2) devise experiments to learn how to move the real numbers
closer to the ideal reflected in the business plan.

innovation accounting works in three steps

1. use a MVP to establish real data on where the company is right now. without a clear
picture of the present, there’s no way you’re getting to the future.
2. tune the growth engine from the baseline (present) to the ideal. this may take many
attempts.
3. after the company has made all the micro changes and product optimisations it can, you
reach a decision point: pivot or persevere.
 a. the sign of a successful pivot is that the micro engine tunings are much more
successful in moving the (new) baseline towards the (new) ideal than before the
pivot.

establish the baseline

there’s a few different ways of doing this. a startup might create a complete prototype of its
product and offer to sell it through its main marketing channel. this is a singular product which
tests all of the startup’s assumptions and test all baseline metrics simultaneously.

alternatively, a startup might prefer to build one MVP for each assumption and/or metric. for
example, performing a smoke test (direct marketing of a product that does not exist, to see if
potential clients would be interested). this is not enough to test the entire growth model, but it is
useful to get feedback on one assumption (are clients interested in this product - or maybe
pricing, and whatnot) before committing more resources.

An MVP allows a startup to fill in real baseline data in its growth model—conversion rates,
sign-up and trial rates, customer lifetime value, and so on—and this is valuable as the
foundation for learning about customers and their reactions to a product even if that foundation
begins with extremely bad news.

test the riskiest assumptions first. for example if you are a media company selling to advertisers,
there are two basic assumptions - can it capture attention, and can it sell that to advertisers? in
a business in which the advertising rates for a particular segment are well known, it makes more
sense to test the first assumption first. so producing a pilot episode/content to see if there is
attention, rather than advertising sales.

tuning the engine

every initiative a startup undertakes (product development, marketing etc) should be aimed at
improving one of the drivers of its growth model.

For example, a company might spend time improving the design of its product to make it easier
for new customers to use. This presupposes that the activation rate of new customers is a driver
of growth and that its baseline is lower than the company would like. To demonstrate validated
learning, the design changes must improve the activation rate of new customers. If they do not,
the new design should be judged a failure. This is an important rule: a good design is one that
changes customer behavior for the better.

a successful startup should have a clear baseline metric(s), a hypothesis around what will
improve it, and a set of experiments designed to test it. only startups that can iterate through
build-measure-learn at lightning speed can succeed.
if you are experimenting properly, with clear baseline metrics, and you see the horrible numbers
of the MVP not improve, and the ideal laid out in your business plan recede even further, then it
is a sure sign it is time to pivot.

cohort analysis
Instead of looking at cumulative totals or gross numbers
such as total revenue and total number of customers, one
looks at the performance of each group of customers that
comes into contact with the product independently. Each
group is called a cohort. The graph shows the conversion
rates to IMVU of new customers who joined in each
indicated month. Each conversion rate shows the
percentage of customer who registered in that month who
subsequently went on to take the indicated action. Thus,
among all the customers who joined IMVU in February
2005, about 60 percent of them logged in to our product
at least one time.

this model is used a lot in enterprise sales - Lean Startup

uses it in product development as well.

at IMVU, despite a fourfould increase in people who logged on more than five times, people who
paid was stuck at 1% for months. in the early days, they had the assumption that people wanted
this produt to be an addon to IM, so they could speak to existing friends. but with data in hand,
he (head of product) was forced to go to customers and find out why people weren’t paying. and
he couldn’t dismiss customers who ddin’t like the product as ‘not target demographic’ this time
around - since each cohort is an ‘independent’ (insert statistical disclaimers here?) set of
customers. only because of this data was he able to find out that people wanted IMVU to be a
standalone network to make new friends, as opposed to speaking to existing ones online. once
they figured that out, their numbers started improving massively after their experiments.

optimisation versus learning

with startups it is hard to optimise because you do not know what you are optimising for. it is
dangerous, because you could assume your engineering team isn’t working hard enough - while
what’s actually happening is you’re not getting enough customer feedback and your
metrics/assumptions aren’t clear.

vanity metrics
now here’s a similar graph graph as above, but in gross figures.
he also gives the example of a company called
Grockit, an online education startup. started
properly with a MVP (just the founder doing
online classes on WebEx) excellent team with
big vc funding, which was even following agile
development: they would focus on a certain
baseline (eg new acquisitions, retention rate) for
a one month period, presented as ‘user stories’ -
descriptions of features in nontechnical language
so that the engineers would always be focussed
on customers - but the founder was uncertain
that the metrics he was prioritizing were the right
ones (the eternal startup dilemma, apparently).
so they switched to cohort-based metrics, and
launched each new feature as a split-test
experiment (A/B testing).

in Lean Startup, A/B testing is not just marketing,

but part of product development. with each feature, half of existing customers would keep on
using the standard product, the other half would use the new version. then you compare
metrics. amazon is well known for having used this a lot in their early days. in grockit’s case,
split tests revealed that customers did not really want more ways to interact with each other
(although the founding team thought that was really important).

kanban
yeet. grockit split user stories into four states of
development: in backlog, actively being built,
done (feature complete from a technical
standpoint), process of being validated. the
‘buckets’ (states) only could take so many
stories, but a story could only have been
removed had it been validated. validation was
done through comparing split metrics, and
sometimes interviews and surveys as well.

engineers don’t like kanban because it forces them to do nonengineering things, like talking to
customers and conducting split tests. ew talking to people. but that’s why it’s good. Incidentally,
a solid process lays the foundation for a healthy culture, one where ideas are evaluated by merit
and not by job title (once engineers start demanding everything be split tested and question the
hypotheses behind the executives’ user stories, things are going well).
three As of good metrics

actionable
for a report to be considered actionable, it must demonstrate clear cause and effect. take
website hits - that’s a vanity metric. you don’t know what caused them, who they’re from, etc.

accessible
make reports simple and easy to understand. ‘metrics are people too’. this is another reason
why cohort analysis is excellent - it puts the focus on people that exhibit behaviours the
company cares about.

Accessibility also refers to widespread access to the reports. Grockit did this especially well.
Every day their system automatically generated a document containing the latest data for every
single one of their split-test experiments and other leap-of faith metrics. This document was
mailed to every employee of the company: they all always had a fresh copy in their e-mail
in-boxes. The reports were well laid out and easy to read, with each experiment and its results
explained in plain English.

Another way to make reports accessible is to use a technique we developed at IMVU. Instead of
housing the analytics or data in a separate system, our reporting data and its infrastructure were
considered part of the product itself and were owned by the product development team. The
reports were available on our website, accessible to anyone with an employee account. Each
employee could log in to the system at any time, choose from a list of all current and past
experiments, and see a simple one-page summary of the results.

auditable
when a person’s pet project is killed on account of data, they will (all too often, apparently)
challenge the veracity of this data. so the data good metrics produce must also be credible.
Drawing reports from the master data (as opposed to an intermediate system) is a way to
reduce error.

pivot or persevere
Only 5 percent of entrepreneurship is the big idea, the business model, the whiteboard
strategizing, and the splitting up of the spoils. The other 95 percent is the gritty work that is
measured by innovation accounting: product prioritization decisions, deciding which customers
to target or listen to, and having the courage to subject a grand vision to constant testing and
feedback.
whether to pivot or not is the most difficult, time consuming, and often wasteful decision an
entrepreneur faces. Everything that has been discussed so far is a prelude to a seemingly
simple question: are we making sufficient progress to believe that our original strategic
hypothesis is correct, or do we need to make a major change? That change is called a
pivot: a structured course correction designed to test a new fundamental hypothesis about the
product, strategy, and engine of growth.

the land of the living dead

it’s even harder to pivot when you’ve had moderate success (but not enough to live up to the
vision and/or the investors). this is known in silicon valley as the land of the living dead - some
success, but not a lot.

to minimise this, committing to iteration and the build-measure-learn cycle early is key: it’s
easier to pivot the fewer resources (time, money, promises/loyalty) you’ve spent on an idea.
also, identify your leap-of-faith predictions at the start, and make quantitative predictions about
each of them. in this way, you can avoid vanity metrics. also, don’t start doing PR until you have
a working product.

Seasoned entrepreneurs often speak of the runway that their startup has left: the amount of
time remaining in which a startup must either achieve lift-off or fail. This usually is defined as
the
remaining cash in the bank divided by the monthly burn rate, or net drain on that account
balance. when you start to run out of runway, you can cut costs or raise more money. but if you
cut too many costs, you may reduce the speed at which you can move through
build-measure-learn, thereby actually decreasing your runway. so the true measure of runway is
how many pivots a startup has left.

also, when an entrepreneur has an unclear hypothesis, it’s almost impossible to experience
complete failure, and without failure there is usually no impetus to embark on the radical change
a pivot requires. As I mentioned earlier, the failure of the “launch it and see what happens”
approach should now be evident: you will always succeed—in seeing what happens. Except in
rare cases, the early results will be ambiguous, and you won’t know whether to early results will
be ambiguous, and you won’t know whether to pivot or persevere, whether to change direction
or stay the course.

reminder that a pivot is a special kind of change designed to test a new fundamental hypothesis
about the product, business model, and engine of growth. It’s not just any change.
● zoom in pivot: what was considered to be only a single feature becomes the entire
product. eg votizen started as a full social network for registered voters, then pivoted to a
simpler voter contact product, turning digital petitions into physical ones mailed to
congress
 ● zoom out pivot: the reverse. what was considered the whole product becomes a single
feature in a much larger product
● customer segment pivot: the company realises it is building a product that solves a real
problem for real customers, just not the one they were expecting to solve it for.
● customer need pivot: in the process of building a product, you become very acquainted
with customer needs, and realise you’re not really solving a concern they have, but
because you understand those needs you either reposition your product or build
something else entirely. an example is potbelly sandwich shop, which started in 1977 as
an antique store.
● platform pivot: change from an application to a platform, or vice versa. usually the
company starts by building the ‘killer app’, and then expands into a platform. but this
pivot is not set in stone, and may need to be executed multiple times.
● business architecture pivot: geoffrey moore observed that companies generally follow
one of two major business architectures - high margin, low volume (complex systems
model), low margin, high volume (volume operations model). usually b2b and b2c
respectively. in this pivot, a startup switches architecture. an example is google going
mass market with their search engine, back in the day.
● value capture pivot: capturing value is an implicit part of the product hypothesis.
changing monetization strategy is too limiting - in reality, this type of change will have far
reaching consequences on the business
● engine of growth pivot: changing growth strategy to seek faster or more profitable
growth. commonly but not necessarily, this also requires a value capture pivot
● channel pivot: the mechanism by which the company delivers its product to customers is
called the sales/distribution channel. the requirements of the channel often determine
the product’s price, features and competitive landscape. A channel pivot is a recognition
that the same basic solution could be delivered through a different channel with greater
effectiveness, ie going direct to customer and abandoning a complex sales process.
● technology pivot: doing the same thing with a different technology. The only question is
whether the new technology can provide superior price and/or performance compared
with the existing technology. Established companies excel at this, as the value prop
stays the same.

again reminder that the pivot is not just an exhortation to change. Remember, it is a special kind
of structured change designed to test a new fundamental hypothesis about the product,
business model, and engine of growth. It is the heart of the Lean Startup method. It is what
makes the companies that follow Lean Startup resilient in the face of mistakes: if we take a
wrong turn, we have the tools we need to realize it and the agility to find another path.

Part 3 - Accelerate
most question a startup face are not clear-cut. The critical first question for any lean
transformation is: which activities create value and which are a form of waste? value in a startup
is not producing stuff, rather it is validated learning about how to build a sustainable business.
What products do customers really want? How will our business grow? Who is our customer?
Which customers should we listen to and which should we ignore? These are the questions that
need answering as quickly as possible to maximize a startup’s chances of success. That is what
creates value for a startup.

there are once again lessons from lean manufacturing - single batch processes, just in time
delivery. this minimizes excess inventory/waste.

in entrepreneurship, this can be applied to releasing small changes one at a time, and
measuring (objectively) their impact. also, checking defects immediately can translate to
extensive automated tests (continuous deployment). a striking example of continuous
deployment is Wealthfront.

continuous deployment is affecting many industries in various ways: hardware becomign

software (smartphones are basically screens connected to the internet, even cars are becoming
ever more connected), fast production changes, 3d printing and rapid prototyping tools.

The essential lesson is not that everyone should be shipping fifty times per day but that by
reducing batch size, we can get through the Build-Measure-Learn feedback loop more quickly
than our competitors can. The ability to learn faster from customers is the essential competitive
advantage that startups must possess.

resist the urge to work in large batches, even if it seems more efficient. the large batch death
spiral will kill you, because unlike in manufacturing there’s no upper limit to batch size. ‘oh, well,
it’s already been a year since our last release - may as well take another month to add a few
more features, right?’

As soon as we formulate a hypothesis that we want to test, the product development team
should be engineered to design and run this experiment as quickly as possible, using the
smallest batch size that will get the job done. Remember that although we write that will get the
job done. Remember that although we write the feedback loop as Build-Measure-Learn because
the activities happen in that order, our planning really works in the reverse order: we figure out
what we need to learn and then work backwards to see what product will work as an experiment
to get that learning. Thus, it is not the customer, but rather our hypothesis about the customer,
that pulls work from product development and other functions.

it seems i should really read a detailed book about toyota at some point.

Grow
new customers come from the actions of past customers.
1. word of mouth
 2. side effect of product usage, ie luxury goods or viral growth of paypal and facebook
3. through funded advertising. for it to be sustainable this must be funded through revenue
and not investment capital. as long as marginal cost < marginal revenue
4. through repeat purchase or use. some products are designed for repeated purchase ie a
subscription or consumables.

engines of growth help narrow options, figure out which metrics are the important ones to track.
there are 3 main engines.

sticky growth
companies depending on ‘sticky’ growth rely on securing customers and maintaining them
forever. an example might be a startup that seeks to become a definitive marketplace for a
niche, or a database company.
attrition (churn) rate is a very important metric for companies that rely on sticky growth. The
churn rate is defined as the fraction of customers in any period who fail to remain engaged with
the company’s product.

The rules that govern the sticky engine of growth are pretty simple: if the rate of new customer
acquisition exceeds the churn rate, the product will grow. The speed of growth is determined by
what I call the rate of compounding, which is simply the natural growth rate minus the churn
rate.

viral growth
the growth of these companies can be modelled mathematically through the ‘viral coefficient’.
growth depends on how many new customers each existing customer brings with them. a viral
coefficient >1 means exponential growth.

paid growth
companies grow by paying for new customers (through advertising, sales processes, foot traffic
- retail etc).
cost per acquisition is the most important metric here. as long as the CPA is less than each
customer’s lifetime value (the amount of money they will give you over their lifetime, after
variable costs have been deducted), you have a viable growth model.

Thus, the ability to grow in the long term by using the paid engine requires a differentiated ability
to monetize a certain set of customers.

also note that if you’ve clearly defined your engine of growth, this works in tandem with
innovation accounting - you can focus on those metrics that would hinder your engine’s
progress, and track them. for example, a startup attempting a viral growth model need not focus
on sales or advertising, all it should focus its development efforts on is that which might affect
customer behaviour. conversely, a startup relying on a paid engine of growth should develop
sales and advertising functions urgently.

Adapt
bear in mind that you cannot trade quality for speed. toyota used to say ‘stop production now so
that production never has to stop’, in reference to their andon cord. if you rush to deliver your
first MVP and it is truly shitty in the sense that it doesn’t work, there are bugs etc, your second
MVP will be slowed down by this. a startup is not a factory churning out stuff, a startup is a
series of strategic learning experiments. if your first MVP is slowing down learning because it’s
so bad you have to spend a lot of time fixing, you have not succeeded. in other words, the
programmer that goes slow is going fast.

five whys
instead, you should use the toyota five whys approach when evaluating something.

1. Why did the machine stop? (There was an overload and the fuse blew.)
2. Why was there an overload? (The bearing was not sufficiently lubricated.)
3. Why was it not lubricated sufficiently? (The lubrication pump was not pumping sufficiently.)
4. Why was it not pumping sufficiently? (The shaft of the pump was worn and rattling.)
5. Why was the shaft worn out? (There was no strainer attached and metal scrap got in.)

as you can see, stopping at 1 would have delayed this particular problem by a few months at
most. also note that as you move from 1 to 5, the problem becomes increasingly human (from a
blown fuse to somebody forgetting to attach a strainer). this is common in all problems a startup
faces.

now a software example:

1. A new release disabled a feature for customers. Why? Because a particular server failed.
2. Why did the server fail? Because an obscure subsystem was used in the wrong way.
3. Why was it used in the wrong way? The engineer who used it didn’t know how to use it
properly.
4. Why didn’t he know? Because he was never trained.
5. Why wasn’t he trained? Because his manager doesn’t believe in training new engineers
because he and his team are “too busy.”

here’s the crux: here’s how to use Five Whys analysis to build an adaptive organization:
consistently make a proportional investment at each of the five levels of the hierarchy. In other
words, the investment should be smaller when the symptom is minor and larger when the
symptom is more painful. We don’t make large investments in prevention unless we’re coping
with large problems. In the example above, the answer is to fix the server, change the
subsystem to make it less error-prone, educate the engineer, and, yes, have a conversation with
the engineer’s manager

also, the Five Whys approach acts as a natural speed regulator. The more problems you have,
the more you invest in solutions to those problems. As the investments in infrastructure or
process pay off, the severity and number of crises are reduced and the team speeds up again.
With startups in particular, there is a danger that teams will work too fast, trading quality for time
in a way that causes sloppy mistakes. Five Whys prevents that, allowing teams to find their
optimal pace. The Five Whys ties the rate of progress to learning, not just execution. Startup
teams should go through the Five Whys whenever they encounter any kind of failure, including
technical faults, failures to achieve business results, or unexpected changes in customer
behavior.

Coupled with working in small batches, this technique provides the foundation a company needs
to respond quickly to problems as they appear, without overinvesting or overengineering. In
other words, five whys + small batches (incremental approach, continuous deployment) are the
essence of the Lean Startup.

also beware of turning the Five Whys into the Five Blames. make sure everyone is present who
was involved in the issue (those who picked up on it, those who escalated to senior
management) so that the party who is not in the meeting does not become the sacrificial lamb of
everyone’s hatred.

At IMVU we told new hires, “If our production process is so fragile that you can break it on your
very first day of work, shame on us for making it so easy to do so.” If they did manage to break
it, we immediately would have them lead the effort to fi the problem as well as the effort to
prevent the next person from repeating their mistake. For new hires who came from companies
with a very different culture, this was often a stressful initiation, but everyone came through it
with a visceral understanding of our culture.

getting started with five whys

for five whys to work, an environment of mutual trust and empowerment must exist. if it does
not, asking these questions can be too overwhelming. instead, a simplified version exists

1. be tolerant of all mistakes the first time

2. never allow the same mistake to happen twice

remember, most mistakes are caused by flawed systems, not bad people. this system is fine
because it encourages more complex questions to be asked (what types of errors should we
focus on? should we seek to prevent a specific error or categories?). precisely because of this,
eventually it needs to grow into something systematic and comprehensive, like Five Whys.
facing unpleasant truths
especially at the start, this process will expose a lot of flaws in your company. it will degenerate
into five blames sometimes. executive leadership needs to be present in all of these situations,
to continue to promote and sponsor the process of five whys.

also, you can always follow the simple two-rule version in your own work. Whenever something
goes wrong, ask yourself: How could I prevent myself from being in this situation ever again?

Once you are ready to begin, I recommend starting with a narrowly targeted class of symptoms.
Even though your business will have bigger problems that need to be addressed, the higher the
stakes the likelier things degenerate into five blames. so start with something relatively minor so
that the company can get used to it. equally, don’t start with radical changes immediately. pick
something relatively easy at each of the five levels to address, and keep meetings short.
appointing a five whys master is a good idea.

my own reflection: this book was written eight years ago, when lean startup was still relatively
new. now, everyone at JPMorgan has read the book twice. What happens now?

small batch size in software can also mean many github branches (though it doesn’t need to).
there are many interesting ways software companies can do small batch sizes, even in very
delicate industries. for example, financial accounts company intuit - to get the batch size down,
the QuickBooks team built a virtualization system that allowed them to run multiple versions of
QuickBooks on a customer’s computer. The second version could access all the customer’s
data but could not make permanent changes to it. Thus, there was no risk of the new version
corrupting the customer’s data by accident.

also there are always non technical solutions to reduce batch size as well - for example,
quickbooks started to experiment with a subscription model for the most active customers -
downloading updates online meant they could release faster than the traditional boxed desktop
product, which has a lengthy sales cycle.

Innovate
startup teams require these things to innovate:
1. scarce but secure resources
2. independent authority to develop the business
3. a personal stake in the outcome

in large companies, division heads are accustomed to playing politics to secure massive
budgets, and equally accustomed to defending these budgets from incursions from other
departments. losing, say, 10% of your budget isn’t a big deal. in a startup, that can mean death.
startups are simultaneously easier and more demanding - they need much less capital, but it
cannot be tampered with.

too many approvals and handoffs slow down the build-measure-learn cycle, so internal startups
should be able to ship full prototypes and products as they see fit.

personal stakes need not be financial, especially at NGOs and the government. even having
your name on the door will do. that makes it much more personal/real.

established companies can create ‘innovation sandboxes’ to promote innovation (isolating

certain team for a certain time with an impact only on certain customers, for the duration of an
experiment).

in summary: Whenever possible, the innovation team should be crossfunctional and have a
clear team leader, like the Toyota shusa. It should be empowered to build, market, and deploy
products or features in the sandbox without prior approval. It should be required to report on the
success or failure of those efforts by using standard actionable metrics and innovation
accounting.

In the twenty-first century, we face a new set of problems that Taylor could not have imagined.
Our productive capacity greatly exceeds our ability to know what to build.

tldr
As we’ve seen, too many innovation teams engage in success theater,
selectively finding data that support their vision rather than exposing the
elements of the vision to true experiments, or, even worse, staying in
stealth mode to create a data-free zone for unlimited “experimentation” that
is devoid of customer feedback or external accountability of any kind.

Anytime a team attempts to demonstrate cause and effect by placing

highlights on a graph of gross metrics, it is engaging in pseudoscience.
How do we know that the proposed cause and effect is true? Anytime a
team attempts to justify its failures by resorting to learning as an excuse, it
is engaged in pseudoscience as well. If learning has taken place in one
iteration cycle, let us demonstrate it by turning it into validated learning in
the next cycle. Only by building a model of customer behavior and then
showing our ability to use our product or service to change it over time can
we establish real facts about the validity of our vision.