Build-measure-learn: how to run the loop from The Lean Startup

Who should use it, and when not to

Use it when a team is about to spend real time or money on something new and the core assumptions are untested: a new product, a new service line, a new market, a pricing model. It also helps a product that has been live for a while, when numbers are flat and the team argues about whether to change direction.

It is not the tool for tuning a product that already works and has steady traffic. Comparing two versions of a sign-up page with thousands of visitors a week is an A/B test, with sample sizes and significance; use an A/B test plan for that. A rough rule: if you don't yet know whether anyone wants it, run the loop; if you know they want it and you're choosing between versions, run an A/B test.

The core concepts

Leap-of-faith assumptions. The beliefs everything else depends on. Ries puts two above the rest. The value hypothesis: customers get real value once they use the product. The growth hypothesis: new customers will find it and come. Test the riskiest one first.

Minimum viable product (MVP). Ries defines it as the version that gets a team the most validated learning about customers for the least effort. It can be a landing page with a price, a service done by hand behind a simple front end, a demo video, or a pre-order. Anything that doesn't feed the metric gets cut.

Validated learning. Progress measured by evidence of what customers do, not by features shipped. "People said they'd use it" is opinion. "Eleven of 40 trial users paid" is learning.

Innovation accounting. A way to measure progress when revenue is still near zero. Ries lays out three milestones: use an MVP to get a baseline, tune the product to move the numbers toward the goal, then decide to pivot or persevere. It works only with actionable metrics, such as per-cohort conversion. Vanity metrics like total sign-ups rise whatever you do.

Plan in reverse. The book's advice: although the work runs build, measure, learn, the planning runs the other way. Decide what you need to learn, then what to measure, and only then what to build.

Pivot or persevere. A pivot is a structured change of course that tests a new basic hypothesis: a different customer, problem, channel, price, or solution. Ries suggests holding the decision as a regular, scheduled meeting, so flat numbers don't drift on for months.

A worked mini example

This is an illustration. A small bookkeeping firm wants to sell a flat-fee "books closed in 3 days" service to online shops with 1 to 10 staff, at $240 a month. Eight shop owners were interviewed; six said month-end bookkeeping is painful. The firm can spend three weeks and $600 on a test.

Note what isn't on the card: software, onboarding automation, or three pricing tiers. None of it answers the question.

How to run one loop

1. List the leap-of-faith assumptions. Write every belief the plan depends on and tag each as a value hypothesis, a growth hypothesis, or another make-or-break belief.
2. Pick the riskiest one. Choose the assumption most likely to be false and most costly if it is, and test only that one in this loop.
3. Write what you need to learn. State it as a falsifiable sentence about customer behavior, such as paying, signing up, returning, or referring, not about opinions.
4. Set the metric and threshold before you build. Choose one per-cohort behavior metric and the pass line that would make the business work, and write down why that number.
5. Build the smallest test. Make the cheapest MVP that produces the metric with real customers, and list what you are leaving out.
6. Run it to the decision date. Fix the end date and minimum number of customers in advance, and don't move the threshold once results come in.
7. Decide pivot or persevere. Persevere if the metric clears the bar, pivot on a named part of the plan if it falls clearly short, and pick the next assumption to test.

The build-measure-learn skill turns your idea into this one-page card.

Common mistakes

• Building first and choosing metrics later. Then any result looks like progress. Plan in reverse.
• An MVP that is just a smaller product. If it takes months, it is not the smallest test.
• Counting compliments. Survey "yes" answers cost the customer nothing. Ask for money, time, or a commitment.
• Moving the bar. Changing the threshold after the data arrives makes every test pass.
• Persevering by default. Without a scheduled decision, teams keep going on flat numbers.

Once the idea holds, the next question is which customers to win first; see choosing a beachhead segment. To learn why customers switch, see the switch interview.

Where it comes from

Eric Ries set out the method in The Lean Startup (Crown Business, 2011), building on years of writing on his blog, Startup Lessons Learned, where he described the MVP in 2009. The book's principles include validated learning, innovation accounting, and the build-measure-learn loop; leap-of-faith assumptions, innovation accounting's three milestones, and planning the loop in reverse all come from it. In a 2013 Harvard Business Review article, Steve Blank described the lean start-up as a replacement for the old routine of writing a business plan, pitching investors, and then selling. The Japanese edition, リーン・スタートアップ, was published by Nikkei BP in 2012.

This page is an independent summary in our own words. It is not affiliated with or endorsed by Eric Ries, his company, or the publisher. "Lean Startup" is used here only to name the book. More explainers are in the strategy books index.

Questions

What is the build-measure-learn loop?

It is the core cycle in Eric Ries's The Lean Startup: turn an idea into the smallest product that real customers can use, measure how they respond, and learn whether to keep going (persevere) or change a basic part of the plan (pivot).

What is an MVP?

Ries defines the minimum viable product as the version of a new product that lets a team collect the most validated learning about customers with the least effort. It is a test, not a small version of the final product.

How is it different from an A/B test?

An A/B test compares two versions of something that already has enough traffic to split. The build-measure-learn loop tests whether a new idea's basic assumptions hold at all, often before there is a product.