The Lean Startup is a method for building companies under uncertainty, published in 2011 by the entrepreneur Eric Ries. It holds that a startup exists to learn what customers want as quickly and cheaply as possible, through repeated build-measure-learn cycles that test one risky assumption at a time. Its central instrument is the minimum viable product, defined as an experiment rather than a small product.
Across the entire pre-launch and early-post-launch arc. The framework defines what counts as evidence (validated learning), what to build first (MVP), how to measure (cohort analysis vs vanity), and when to change direction (pivot vs persevere). It's a general discipline more than a single tool.
What the Lean Startup is
The Lean Startup is a method for building companies under conditions of extreme uncertainty. It was published in 2011 by the entrepreneur Eric Ries, drawing on lean manufacturing, customer development work by Steve Blank, and agile software practice.
Its central claim is that a startup is not a small version of a large company but an organization whose purpose is to learn what customers want as quickly and cheaply as possible. The unit of progress is validated learning: evidence about a risky assumption, obtained from real customer behaviour rather than from opinion or analysis.
The method runs as a loop. Build the smallest thing that tests one assumption, measure what happens, learn whether the assumption held, and decide whether to persevere or pivot. Its best-known instrument, the minimum viable product, is defined as an experiment rather than as a small product, which is the distinction most teams lose.
- Name the riskiest assumption Leap of faith
The belief that, if wrong, ends the business. Usually about value or about growth, and usually unexamined.
- Run the build-measure-learn loop Minimise total time
The goal is to get through the loop fast, not to build fast. Most delay is in deciding what to measure.
- Measure with cohorts, not totals Actionable vs vanity
A cumulative number rises even while the product fails. An actionable metric can fall, which is what makes it evidence.
- Persevere, or pick one of ten pivots Not a binary
Ries names ten. The question is not whether to change but which single element to change while keeping the rest.
Why it matters
When The Lean Startup shipped in 2011 it changed how early-stage teams talked about their work. Within a few years “lean” had become a cargo cult: founders called underbuilt products lean, called underfunded teams lean, and skipped the discipline that gave the word meaning.
The discipline is unglamorous and it is the whole method. Name the assumption that would kill you. Build only enough to test it. Measure with cohorts so the answer can be no. Most teams do the first and third steps informally and the second step enthusiastically, which produces a lot of shipping and very little learning.
“We are being lean” has become, in most rooms, a sentence about the budget. Ries wrote a book about how to be wrong quickly, and the industry heard a book about spending less. That is a very ordinary way for a good idea to be digested.
There is a finding in ShipFit’s corpus that sits uncomfortably beside all of this, and it is the strongest thing in the dataset. Whether a founder finishes a nine-stage validation process is predicted, with almost embarrassing accuracy, by what they were told at stage one.
Founders told their idea showed a strong signal on question one were thirty-nine times more likely to finish all nine than founders told to go and find out.
- Strong signal
- n = 57, mean stage 7.75 78% finished
- Worth exploring
- n = 29, mean stage 4.79 37% finished
- Promising, needs focus
- n = 352, mean stage 3.77 8% finished
- Let us find out
- n = 82, mean stage 2.84 2% finished
Sample: n = 548 ideas with a first-question verdict
What it does not say: An honest alternative reading: founders may simply abandon ideas the tool was lukewarm about, which is arguably the tool working rather than a prediction.
| Rises no matter what | Can tell you you are wrong | Why the swap |
|---|---|---|
| Total registered users | Cohort retention at week 4 | Cumulative totals only ever go up, including while the product is failing. |
| Pageviews | Activation rate per cohort | Traffic measures marketing spend, not whether the product worked. |
| Total revenue | Revenue per cohort over time | Aggregate revenue hides that each new cohort is worth less than the last. |
| Number of downloads | Share still active on day 30 | A download is a click. Activity is evidence. |
When to run it
- You are about to commit months of build to an assumption nobody has tested.
- Growth stalled and the team disagrees about why.
- You have shipped steadily for two quarters and cannot name what you learned.
- You are considering a major change of direction and want evidence rather than a vote.
- Your metrics all go up and your revenue does not.
- You could find out by asking ten people this week. Use The Mom Test →
- The question is what to charge. Use Van Westendorp →
- You need to cut a feature list down to a shippable release. Use MoSCoW →
- You want to know whether you have product-market fit at all. Use the Superhuman PMF engine →
What an MVP actually is
A minimum viable product is the smallest thing that produces validated learning about one assumption. It is an experiment, and it is judged by what it teaches rather than by what it does.
That definition rules out most things called MVPs. A stripped-down version of the real product, shipped to see how it goes, is not an experiment because no assumption was named in advance and no result would have counted as a failure. It is version 0.8.
The forms that qualify are often not software at all: a landing page that measures whether anyone will give you an email address, a concierge service you deliver by hand, a Wizard-of-Oz product with a person behind the curtain. Each isolates one belief and puts a number on it.
The three engines of growth
The loop tells you to measure. The engine tells you which number.
New customers arrive faster than existing ones leave.
Churn rate, against the rate of new acquisition
Chasing acquisition while churn quietly cancels it out.
Using the product exposes it to new users as a side effect.
Viral coefficient. Above 1.0 and growth compounds on its own.
Confusing word of mouth with a viral loop built into normal use.
Revenue per customer exceeds the cost of acquiring them, and the surplus buys more.
LTV against CAC, and how fast CAC is recovered
Spending against a lifetime value nobody has observed yet.
Pick one. Ries is explicit that a startup running all three at once is running none of them well, because each demands a different product decision and a different definition of a good week.
Pivot is not a binary
“Pivot or persevere” is the phrase everyone remembers, and it makes the decision sound like a yes or no about changing direction. Ries names ten pivots, and the useful question is which single element to change while keeping the validated learning from everything else.
- Zoom-in
One feature becomes the whole product and everything else is cut.
Signal: Usage is concentrated in a corner of the product nobody planned for.
- Zoom-out
The whole product becomes one feature of something larger.
Signal: People like it and it is not enough on its own to be worth buying.
- Customer segment
Same product, a different buyer than the one you built it for.
Signal: The wrong people love it and the right people are indifferent.
- Customer need
Same buyer, a different problem you found while talking to them.
Signal: The relationship is good and the problem you picked was not the painful one.
- Platform
Application becomes platform, or platform retreats to a single application.
Signal: Users are building on top of you, or nobody is.
- Business architecture
High margin and low volume, or low margin and high volume. Swap.
Signal: Your cost of sale does not match the price the market accepts.
- Value capture
A different way of charging for the same value.
Signal: People use it heavily and the revenue model does not track the usage.
- Engine of growth
Switch between the viral, sticky and paid engines.
Signal: The growth loop you assumed is not the one actually turning.
- Channel
A different way of reaching the same buyer.
Signal: Conversion is fine and nobody arrives.
- Technology
Same problem, same buyer, a fundamentally different technical approach.
Signal: The solution works and cannot be delivered at a viable cost.
A pivot keeps one foot planted. Every one of these changes a single element and holds the validated learning from everything else, which is what separates it from starting over. A team that changes the buyer, the problem, the pricing and the channel at once has not pivoted; it has begun a different company with the same bank account.
Lean validation in practice: the Dropbox video
The most efficient assumption test in the canon, and a useful reminder of how narrow a good test is.
Dropbox · 2007 to 2008
Tested demand with a three-minute video, before the product could do what the video showed.
File sync was hard to explain and harder to demo, and building enough of it to demo credibly was months of work against an unproven assumption. Drew Houston instead made a screencast of the product working, seeded with in-jokes aimed squarely at the Digg and Hacker News audience, and posted it.
The waiting list is reported to have gone from around 5,000 to 75,000 overnight. No product shipped that night.
It is worth being precise about what this tested. It did not prove people would pay, or that sync could be built reliably, or that the business would work. It tested one assumption, the one that would have been most expensive to be wrong about, and it tested it in the cheapest currency available.
- Waiting list before
- ~5,000
- Waiting list after
- ~75,000
- Working product at time of test
- partial
What it shows: A minimum viable product is not a small version of the product. It is the smallest thing that returns a verdict on the assumption most likely to kill you.
The Lean Startup vs the alternatives
How do I find out whether this works, cheaply and before it is too late?
Gives you: Validated learning, and a persevere-or-pivot decision
How do I systematically discover and validate a customer?
Gives you: Blank’s four-step process. Lean Startup builds directly on it
How do I generate solutions worth testing in the first place?
Gives you: Divergent options. Complementary; Lean is weak at generating ideas
How does the team build in short increments?
Gives you: Shipping cadence. Often confused with Lean, and answers a different question
When it won’t help you
- It optimizes; it does not originate
The loop is excellent at improving a hypothesis and says nothing about where good hypotheses come from. Run on a mediocre idea it produces a well-validated mediocre product.
Instead: Pair it with something generative. Jobs to be Done and Blue Ocean both have opinions about where to look.
- Iteration is not the same as learning
The most common cargo-cult version: ship weekly, call it lean, never name an assumption in advance so no result could ever have counted as a failure. This is motion with a changelog attached.
Instead: Write the assumption and the number that would disprove it before you build, not after you see the data.
- Short loops bias toward small questions
The method rewards what can be tested this sprint, which quietly deprioritises anything requiring a long build or a slow market. Some real businesses cannot be validated in two-week increments.
Instead: Match the loop length to the risk. Some assumptions need a quarter, and running them as a fortnight tests something else.
- It assumes you can measure the thing that matters
In long sales cycles, regulated markets or genuinely new categories, the feedback signal arrives long after the decision, and cohort data on a nine-month cycle is not a loop.
Instead: Use leading indicators you have validated against the lagging outcome, and be honest that they are proxies.
ShipFit and the Lean Startup loop
ShipFit front-loads the part of the loop that costs the most when skipped. Stage 1 forces the leap-of-faith assumption into the open, and Stages 2 to 7 are the validation work that tests it before you write code: buyer at Stage 2, pain at Stage 3, solution approach at Stage 4, MVP scope at Stage 5, pricing at Stage 6, demand proof at Stage 7. The output is an MVP defined as an experiment with a named hypothesis attached, rather than a feature list with the word minimum in front of it.
Where this sits in the sequence
The loop is where a scoped release goes to find out whether it worked.
- Jobs to be Done
Establish what the release has to accomplish for the buyer.
- MoSCoW
Cut the candidate list to what the version cannot ship without.
- ICE scoring
Sequence what survived the cut.
- Lean validation
Ship it as an experiment and measure what came back.
You are here
Further reading
- Eric Ries, The Lean Startup (2011). The source.
- Eric Ries, The Startup Way (2017). The same method applied inside large organizations.
- Steve Blank, The Four Steps to the Epiphany (2005). The customer development work Lean Startup builds on.
- The Mom Test. Cheaper than an experiment when the question is answerable by asking.
- Superhuman PMF engine. What to run when the question is whether you have product-market fit at all.
- MoSCoW. How to cut a candidate list down to something a loop can actually test.
- ICE scoring. How to sequence the experiments once you have more than you can run.
- CAC / LTV ratio calculator. The arithmetic behind the paid engine.
How to apply The Lean Startup
- 1
Write down the leap-of-faith assumption
The belief that, if wrong, ends the business. Usually one about value (people want this) and one about growth (we can reach them repeatably). Naming it before you build is what makes the next steps an experiment rather than a release.
- 2
Define what result would prove you wrong
Pick the number and the threshold in advance. An experiment that cannot fail is not an experiment, and deciding what counts as failure after seeing the data is the single most common way this method gets hollowed out.
- 3
Build the smallest thing that tests it
An MVP is judged by what it teaches, not by what it does. A landing page, a concierge service delivered by hand, or a Wizard-of-Oz product with a person behind the curtain all qualify. A stripped-down version of the real product usually does not, because no assumption was isolated.
- 4
Measure with cohorts, never with totals
Cumulative numbers rise even while the product is failing, because new users keep arriving. Cohort retention exposes what total signups conceal. If your metric cannot go down, it cannot tell you that you are wrong.
- 5
Decide: persevere, or which of the ten pivots
Ries names ten, each changing a different element: zoom-in, zoom-out, customer segment, customer need, platform, business architecture, value capture, engine of growth, channel, technology. A pivot keeps one foot planted and preserves the learning from everything else. Changing all of them at once is not a pivot.
- 6
Pick one engine of growth and measure against it
Sticky, viral or paid. The engine decides which number matters, so 'what is our north star metric' is unanswerable until you have chosen. Running all three at once means running none of them well.
Common mistakes
- **Calling everything an MVP.** A bug-fixed early version of the real product isn't an MVP; it's V1.3 of the product. An MVP is a scoped experiment with a clear hypothesis. Most 'MVPs' founders ship are not MVPs.
- **Measuring with vanity metrics.** Total signups goes up over time even when the product is failing because new users keep coming. Cohort retention exposes the failure. Use cohort metrics or you're flying blind.
- **Pivoting too soon or too late.** Too soon: you change direction before the loop has produced clear signal (typically 2-3 iterations). Too late: you keep iterating on a clearly-failing hypothesis because changing direction is emotionally hard. Both fail.
- Misreading 'lean' as 'cheap.' Lean means 'low-waste,' not 'low-investment.' A lean startup might raise $5M to test the leap of faith. The cost is the deliberate, evidence-generating discipline, not the spend total.
- **Treating the framework as a sequential checklist.** Build-Measure-Learn is a loop, not a waterfall. Most actual lean startup work runs the loop concurrently across multiple hypotheses.
How ShipFit operationalizes this
ShipFit operationalizes the Lean Startup loop. Stage 1 (Worth Building?) asks you to name the leap-of-faith assumption underlying the idea. Stages 2-7 are the validation work that tests it before you write code: buyer (2), pain (3), solution approach (4), MVP scope (5), pricing (6), demand proof (7). Stage 7 produces a Smoke test plan and Pre-sales playbook so the behavioral evidence runs before the build commitment, not after.
ShipFit runs 55 frameworks across 9 decision stages
The Lean Startup is one tool in a bigger toolkit. The full library covers market sizing, buyer discovery, MVP scoping, pricing, and launch.
The Mom Test
Q3Rob Fitzpatrick
Validation question methodology, real interviews, not theater
Jobs-to-be-Done
Q2-Q4Clayton Christensen
Functional, social, and emotional jobs your product fulfills
7 Powers
Q4Hamilton Helmer
Strategic moats: Scale, Network, Counter-positioning, Switching, Brand, Cornered Resource, Process
Van Westendorp PSM
Q6Feature-weighted price sensitivity analysis without guessing
Blue Ocean Strategy
Q4Kim & Mauborgne
ERRC framework: Eliminate, Reduce, Raise, Create
Fake Door Testing
Q7Pre-build behavioral validation with landing pages and apology modals
+ 49 more: TAM/SAM/SOM Analysis, Porter's Five Forces, Market Timing Analysis, Unit Economics (LTV/CAC)...
Frequently asked questions
What is the Lean Startup framework?
What is an MVP?
What are validated learning and vanity metrics?
What is a pivot in Lean Startup terminology?
How is Lean Startup different from agile development?
Is Lean Startup still relevant in 2026?
What books pair well with The Lean Startup?
What are the ten types of pivot?
What are the three engines of growth?
What is the difference between a vanity metric and an actionable one?
Is an MVP just a version 1 with fewer features?
What is validated learning?
Keep exploring
The 9-step playbook from market verdict to ship-ready spec.
Rahul Vohra's method for measuring product-market fit and improving it: the 40% benchmark, the survey mechanics that make it comparable, and the roadmap it produces.
The Mom Test is Rob Fitzpatrick's framework for customer interviews that generate real signal. Not praise. Three rules, applied step-by-step, with examples.
Most product launches fail not because the product was bad, but because the launch was a list of channels nobody mapped to a buyer. Here is the template that fixes that.
Default-prompted AI is a slop machine: agreeable, plausible-sounding, useless for validating an idea. Here's how to use AI for the parts where it actually adds signal, and where to keep it out of the way.
How many units a month before the math stops bleeding?
Three honest tiers. DIY: $0-200 in books + 60-100 hours of your time over 6-10 weeks. ShipFit: $5 for a Quick Take, $10 for a full playbook, $19-99/mo subscription. Strategy consultant: $5,000-25,000 for a 4-8 week engagement. Hidden cost across all three: ~$50-200 for buyer-interview coffee. The number to compare against: the $20K-80K it costs to skip validation and ship the wrong thing.
SaaS idea validation that pressure-tests your ICP, pricing model, and retention before you build. ShipFit forces 9 decisions in ~20 minutes. Start free.
IdeaProof scores your idea in 120 seconds and bundles brand assets, AI ads and landing pages. ShipFit forces 9 sequential decisions with live G2 / Trustpilot / Reddit signal, Van Westendorp pricing and exports to 7 coding tools. A score is an opinion. A playbook is a plan.
Ready to make your next product a success?
9 decisions between your idea and a product worth building.