Lean Startup & Build-Measure-Learn · Eric Ries' Validated Learning Loop

The core argument

Traditional product planning treats the product as a known thing — you write a spec, you build it, you ship it. Lean Startup argues that for any genuinely new product, the spec is a guess. The right unit of work is not a feature but an experiment that validates or disproves a hypothesis about customer behavior.

The work then becomes: identify the riskiest hypothesis, design the cheapest experiment to test it, run it, learn from the result, and decide whether to persevere or pivot. Repeat.

The Build-Measure-Learn loop

The loop has three stages, each with its own discipline:

1. Build. Build the minimum viable product (MVP) that lets you test the hypothesis. Not the minimum lovable product, not the minimum viable feature — the minimum experiment.
2. Measure. Define the actionable metrics that will tell you whether the hypothesis was right. Vanity metrics (cumulative signups, page views) don't count.
3. Learn. Make the persevere-or-pivot decision based on what you measured. Document what you learned regardless.

The cycle time matters. A team that runs the loop weekly learns 50 things a year. A team that runs it quarterly learns 4. Compounding learning is the lean advantage.

The MVP problem

"Minimum Viable Product" has become one of the most abused terms in product work. Ries' original definition is narrow: the smallest thing that lets you run a meaningful experiment. In practice, most "MVPs" we see are scoped-down versions of the final product, missing the experiment-design aspect entirely.

A real MVP could be a landing page that asks for emails (test demand). It could be a concierge service where the founder manually does what the product will eventually automate (test workflow). It could be a Wizard-of-Oz prototype where the front-end is real but the back-end is humans (test usage patterns). None of these require building the actual product.