In one sentence: Eric Ries argues that a startup is an institution operating under extreme uncertainty rather than a small version of a big company, that entrepreneurship is a form of management, and that the only honest measure of progress is validated learning about which parts of the plan are wrong.
At a Glance
Author: Eric Ries
First published: 2011 (Crown Business)
Category: Business / Entrepreneurship
Length: about 336 pages, about 84,000 words (Crown Business hardcover)
ISBN-13: 978-0-307-88789-4 (Crown Business hardcover)
Summary reading time: about 12 minutes
Book reading time: about 5.5 hours
Notable adaptations: no film or series, though the book launched the Lean Startup conference and a worldwide network of meetup groups, and Ries followed it with The Startup Way in 2017
Content note: the figures quoted throughout are as of 2011, and several of the companies held up here as successes have since been acquired, pivoted again or closed. Ries discloses equity or advisory relationships in most of the startups he uses as evidence, which the analysis section takes up. All of the book’s charts and diagrams, including the build-measure-learn loop itself, are absent from the source edition used here, so this summary describes them in words.
Ries got here the hard way. His first company let college students build profiles to share with employers, and nobody wanted it. In 2004 he cofounded IMVU, a 3D avatar chat product, and the mistakes he made there as chief technology officer supply most of the book’s evidence. The lineage is borrowed and credited: lean manufacturing, Steve Blank’s customer development, agile practice, Christensen on disruption. It was a blog and a movement before it was a book, which explains the evangelical register.
Read it if you are building something whose customers you cannot yet name, or if you must hold a new product group accountable without a forecast. Skip it if you want fundraising tactics, or if you know Blank’s work and want evidence rather than new vocabulary.
The Big Idea
Ries defines a startup by its conditions rather than its size, and what the definition leaves out is the point: nothing about headcount, industry or funding. A team of five inside Intuit qualifies, while an exact clone of a proven business does not, since its success turns on execution. From this follows the claim that irritates some readers: entrepreneurship is a kind of management, and the absence of management is chaos rather than freedom.
If a startup cannot forecast, what counts as a good week? Ries answers with validated learning, defined as demonstrating empirically that a team has discovered valuable truths about its present and future business prospects. A startup does not know who its customer is, so the only defensible unit of value is knowledge, and what fails to produce it is waste. IMVU is his proof. The team spent 180 days building an instant-messaging add-on meant to ride network effects, launched it, and watched nothing happen. Interviews showed teenagers would not invite their friends and did not mind running several IM clients at once, so he threw away interoperability code for more than a dozen networks. The code was waste. The conversations were progress.
The machinery is the build-measure-learn loop. Turn an idea into a product, measure how customers behave, learn whether to pivot or persevere, begin again. Planning runs backwards, from what you need to learn to the smallest build that shows it. What matters is not the speed of any one stage but the total time around the loop, since that sets how many wrong assumptions a company can find before the money runs out.
Key Ideas
1. Find the leap-of-faith assumptions and test those first
Every plan rests on two assumptions that no amount of discussion can settle. The value hypothesis asks whether the product delivers value once in use, measured by whether people come back. The growth hypothesis asks how new customers will discover it. Ries reports that both tend to hide inside analogies, where a confident comparison to somebody else’s success smuggles in an untested premise. His answer, borrowed from Toyota as genchi gembutsu, is to go and see for yourself, though early contact yields no answers, only a sense of which assumption is most dangerous.
2. The minimum viable product is the cheapest full turn of the loop
Ries defines it as that version of the product which enables a full turn of the loop with a minimum of effort and the least development time, and he stresses that it is not necessarily the smallest product imaginable. Drew Houston’s demonstration video took the Dropbox waiting list from 5,000 to 75,000 overnight, before the software worked. Food on the Table ran a concierge version for a single customer, with a hand-delivered packet and a 9.95 dollar weekly cheque, automated only when the founders ran out of hours.
3. Innovation accounting, and cohorts instead of vanity metrics
Innovation accounting runs in three learning milestones: establish a baseline with a minimum viable product, tune the engine toward the ideal, then pivot or persevere. Metrics must be actionable, accessible and auditable, which disqualifies the gross totals Ries calls vanity metrics, since a rising total can hide a product getting worse. At IMVU, five dollars a day on Google AdWords at roughly five cents a click bought a daily report card. Cohort analysis, tracking each intake of customers separately, showed repeat use climbing from under 5 percent to almost 20 percent while paying customers stayed near 1 percent. Straight C’s, he calls it. The gross graph looked like victory.
4. A pivot is a tested hypothesis, not a change of mind
A pivot, in his definition, is a structured course correction designed to test a new fundamental hypothesis about the product, strategy and engine of growth, and the word, he says, is often misused as a synonym for change. He catalogues ten, from zoom-in and customer segment through platform and channel. Votizen is the instructive case: David Binetti’s first minimum viable product cost about 1,200 dollars and three months, and his pivots then ran eight months, four, three and one. The acceleration is the lesson, and because he had stated quantitative predictions for each leap of faith he could not declare victory on gross numbers, escaping what Ries calls the land of the living dead. Ries admits missing a pivot of his own: at over a million dollars a month and twenty million avatars, IMVU flatlined as the early-adopter market ran out, and a customer segment pivot took revenue past 25 million dollars a year by 2009.
5. Small batches, because problems surface on the first envelope
In a demonstration borrowed from Womack and Jones, a father stuffing envelopes one at a time beats two children who fold every letter first, mostly because a letter that does not fit is found on the first attempt rather than the hundredth. Post-war Toyota could not afford large machines, so Ohno and Shingo attacked changeover time instead, and Shingo’s single-minute exchange of die cut it from hours to under ten minutes. IMVU’s software equivalent released about fifty changes a day behind an automated immune system that pulled a bad change, notified the team and blocked deployment pending root-cause analysis. The lesson, Ries says, is batch size and not the number fifty.
6. Growth runs on one of three engines
Sustainable growth, on his account, obeys one rule: new customers come from the actions of past customers. The sticky engine measures churn against the natural rate of new-customer acquisition. A collectibles marketplace with 61 percent retention and 39 percent new-customer growth was compounding at 0.02 percent. The viral engine is measured by the viral coefficient, and Hotmail’s one-line signature carried it to 12 million users in eighteen months. The paid engine sets lifetime value against cost per acquisition, where scale is irrelevant to the rate: a business earning one dollar per customer at an 80 cent acquisition cost grows exactly as fast as one earning 100,000 dollars at 80,000 dollars.
7. Building an organization that adapts, including inside a big company
Ries’s instrument for the long run is the five whys, Ohno’s practice of asking why until a fault’s real origin appears, paired with proportional investment, so a minor symptom earns the first hour of a remedy rather than the whole eight-week programme. He ran it at IMVU on a server failure and reached a gap in training. The pitfall he names is the five blames, countered by putting everyone connected to the problem in the room and having the senior-most person take the blame first. Inside established firms he prescribes scarce but secure resources, independent development authority, a personal stake in the outcome, and an innovation sandbox where a bounded slice of customers can be split-tested.
Context and Analysis
What has held up best is not the branding but the measurement. Cohort analysis against gross totals is a real and portable point. The scheduled pivot or persevere meeting is a good institutional device because it moves the hardest decision away from the moment of emotion. Treating learning rather than shipping as the unit of progress reframes a great deal of invisible waste. And the vocabulary earns its keep for an unglamorous reason: it gives a team a shared way to disagree about what they actually know. The legacy shows up in accelerators, corporate innovation groups and government digital teams.
The evidence is weaker than the argument deserves, in a way the book should have caught. Ries opens by attacking selection bias, then argues almost entirely from IMVU, Dropbox, Intuit, Food on the Table, Votizen, Grockit, Aardvark and Wealthfront, which his own disclosures page lists as companies in which he held equity or an advisory role. A method argued from the firms its author backed has not been tested against the ones he did not, and no startup appears that ran the loop faithfully and died anyway. The minimum viable product stretches so wide, from a video to a concierge service to a Wizard of Oz front end, that a concept which cannot be violated cannot be tested. Validated learning carries little statistical rigour, and a dozen early adopters can send a team in precisely the wrong direction. The pivot catalogue labels changes after the fact rather than predicting them, and the engine rules, including the instruction to run only one, are asserted. The small-batch case rests on a manufacturing analogy and IMVU’s release cadence, which fits regulated or heavily capitalized products far less well, with the patent, brand and quality objections waved aside as speed bumps. The lineage is credited but largely repackaged, and the two schools he argues against are softer than any serious opponent holds. The epilogue, with Taylor and a proposed long-term stock exchange, offers no evidence.
The sharpest argument against this book is Zero to One, where Peter Thiel holds that valuable companies come from a contrarian secret pursued through a bold plan, and that iterating toward product-market fit produces incrementalism. A founder should sit with both: Ries is strongest where the customer is unknown and being wrong is survivable, Thiel where the thesis rests on a commitment that cannot be tested in pieces. Running underneath the final chapters is The Innovator’s Dilemma, one of Ries’s acknowledged sources, which explains why well-run companies rationally decline the small, low-margin markets that eventually unseat them. The innovation sandbox is his attempt to solve Christensen’s problem from inside.
What It Teaches
Six things to take to work on Monday.
1. Write down your two leap-of-faith assumptions, and name the metric that falsifies each. 2. Build the smallest thing that produces the learning you need, even a video or a person behind a curtain. 3. Report by cohort, and treat a metric that only ever rises as a warning. 4. State a quantitative prediction before each experiment, so a weak result cannot be rewritten as success. 5. Put the pivot or persevere meeting in the calendar in advance, with the full optimization history in the room, not a highlight. 6. When something breaks, ask why five times, and invest in prevention proportionally to the damage.
Memorable Lines
“A startup is a human institution designed to create a new product or service under conditions of extreme uncertainty.” (Ries)
“Startup success can be engineered by following the right process, which means it can be learned, which means it can be taught.” (Ries)
“We must learn what customers really want, not what they say they want or what we think they should want.” (Ries)
“If we do not know who the customer is, we do not know what quality is.” (Ries)
“At the root of every seemingly technical problem is a human problem.” (Ries)
“If we stopped wasting people’s time, what would they do with it?” (Ries)
Should You Read the Full Book?
Verdict: Recommended
This summary carries the framework, the definitions as Ries gives them, and the honest limits of the evidence. What only the book delivers is the texture of the cases: Votizen’s metric tables across four pivots, the IMVU cohort graphs read quarter by quarter. Those details show a method being argued rather than announced. The The Lean Startup book page has the full details and where to get a copy.