Chapter 04 of 06
Eric Ries didn't invent a new methodology. He translated Theory of Constraints into startup language. Build-Measure-Learn is constraint exploitation with a feedback loop.
The loop
The minimum experiment that tests your riskiest assumption. Not the minimum product — the minimum learning. Most MVPs are too big because founders optimize for features instead of questions.
TOC equivalent: Exploit — extract learning from the knowledge constraint with existing resources
Innovation accounting, not vanity metrics. Page views, signups, and downloads don't tell you if customers value the thing. Retention, referral, and willingness-to-pay are throughput. Everything else is inventory.
TOC equivalent: Subordinate — measure only what feeds the constraint, ignore the rest
Validated learning: what did you discover that you couldn't have known without running the experiment? This is the only unit of progress that matters in early stage. Not lines of code, not features shipped.
TOC equivalent: Elevate — the constraint shifts after genuine learning. The loop restarts.
The deeper argument
Steve Blank calls this "the unknown unknowns." You don't know who your customer is. You don't know what problem they actually have. You don't know how much they'll pay. This ignorance is the constraint. The MVP is designed to exploit it — cheaply, quickly, before you build the wrong thing expensively. Adding capital to an unknown-unknowns constraint doesn't accelerate learning; it accelerates building the wrong thing faster.
Vanity metrics make you feel good without telling you whether your constraint is being exploited. Page views go up → constraint still unresolved. MRR from retained customers → constraint is moving. Ries's framework says to track "actionable metrics" — metrics that change based on actions you take — instead of metrics that trend up because you added another feature or ran another ad. Goldratt would say: track throughput, not activity.
Ries's pivot/persevere decision maps almost perfectly to Goldratt's Step 5. When the current constraint is fully exploited, you either persist on the same model (persevere) or discover that the constraint has shifted to a different part of the system (pivot). Most pivots are not failures — they're constraint migration. The danger is pivoting before you've fully exploited the current constraint. That's just running away.
Steve Blank's most famous principle: stop theorizing about customers. Go talk to them. This is TOC applied to knowledge work — your assumption about what customers want is the constraint, and the only way to exploit it is with real feedback from real humans. No amount of internal debate resolves a customer discovery constraint. The building is the bottleneck.
The mapping
| Goldratt (TOC) | Ries (Lean) |
|---|---|
| Identify the constraint | Find riskiest assumption |
| Exploit the constraint | Build the MVP |
| Subordinate everything | Measure (not everything) |
| Elevate the constraint | Pivot or persevere |
| Repeat | BML loop |
| Throughput | Validated learning |
"The only way to win is to learn faster than anyone else."
— Eric Ries
The Lean Startup (2011)
"Get out of the building. No business plan survives first contact with a customer."
— Steve Blank
The Four Steps to the Epiphany
Up next
What massive funding actually buys — and why the answer isn't "progress." The case against blitzscaling.
Chapter 05: The VC Distortion →