# Build–Measure–Learn

A feedback loop that turns a risky assumption into the smallest useful test, behavioral evidence, and a consequential product decision.

- Type: framework
- Domain: Strategy & discovery
- Level: Advanced
- Last reviewed: 2026-09-22
- Original: [Soroush’s Product Wiki](<https://wiki.soroush.uk/topic/build-measure-learn/>)

## What it means

Build–Measure–Learn is a loop for reducing uncertainty about a product or business model. A team identifies what it needs to learn, builds the minimum intervention capable of producing credible evidence, measures relevant behavior, and interprets the result before choosing the next action. Although named in build-first order, responsible planning works backward from the learning decision and evidence required.

## Why it matters

The loop reframes progress from producing more software to reducing consequential uncertainty. It can shorten the distance between a strategic belief and observable evidence, limit waste, and help teams discover that the product, audience, channel, or value mechanism should change before a larger commitment becomes difficult to reverse.

## When to use it

Use it when a material assumption can be tested through customer or system behavior and the result could change direction. Do not use rapid looping to evade safety assurance, accessibility work, legal review, or long-term research whose evidence cannot be responsibly compressed.

## How it works

1. Name the decision and the riskiest assumption preventing it.
2. Work backward to define observable evidence, thresholds, and guardrails.
3. Choose the smallest ethical intervention that can generate valid evidence.
4. Instrument behavior and collect qualitative context, not just aggregate activity.
5. Interpret results against the prior prediction and investigate alternative explanations.
6. Decide to persevere, adapt, pivot, pause, or stop; then design the next loop.

## A practical example

A bank believes customers will use automatic savings rules if they feel reversible and understandable. Before building a rules engine, it offers a supervised prototype to a consented cohort, measures rule creation and cancellation, interviews non-users, and monitors complaints. Low adoption traced to income volatility changes the scheduling concept before platform investment.

## Where it breaks

Fast cycles can optimize locally while missing slow harms, seasonality, network effects, or strategic dependencies. A small test may lack ecological validity, and measurement can alter behavior. Learning quality—not cycle count—is the goal; some questions require larger samples, longer observation, or non-experimental research.

## Common mistakes

- Building first and deciding what to learn afterward.
- Equating a minimum viable product with a low-quality miniature product.
- Collecting easy metrics that cannot change the decision.
- Celebrating loop speed while ignoring validity or ethics.

## Visualization

[View the visualization on the website](<https://wiki.soroush.uk/topic/build-measure-learn/#visualization-build-measure-learn>)

## Sources

1. [Why the Lean Start-Up Changes Everything](<https://hbr.org/2013/05/why-the-lean-start-up-changes-everything>) — Harvard Business Review · Management review article
2. [Testing Business Ideas](<https://www.strategyzer.com/library/testing-business-ideas-book>) — Strategyzer · Creator book overview

## Related topics

- [Validated Learning](<https://wiki.soroush.uk/topic/validated-learning/>)
- [Innovation Accounting](<https://wiki.soroush.uk/topic/innovation-accounting/>)
- [Minimum Viable Product](<https://wiki.soroush.uk/topic/minimum-viable-product/>)
