The method

Seven stages of proof, not a one-shot report.

ValidLab walks an idea from a raw hunch to investor-ready through seven stages, each with explicit gate criteria and a separate quality verdict. The method is the founder's own — two decades of lean product validation, refined through MIT executive education. The point is the same throughout: surface what's actually true about your idea, early, while it's still cheap to change your mind.

The seven stages

  1. 01
    diagnose

    You think you have a problem worth solving? Prove it.

  2. 02
    discover

    Talk to real customers. Not your co-founder.

  3. 03
    define

    Strip it down. Build less. Way less.

  4. 04
    validate

    Does anyone actually want to pay for this?

  5. 05
    ignite

    Light the fire. One channel. One target market.

  6. 06
    deploy

    Systems, not heroics. Make it repeatable.

  7. 07
    dominate

    Scale smart — only when the numbers say so.

What makes the score honest

Gates and quality are different signals

A gate is the minimum to advance a stage — did you fill the required fields? Quality is how strong those answers actually are. A stage can pass its gate on thin answers; ValidLab shows both and never lets the first masquerade as the second.

Evidence moves the number, eloquence doesn't

A traction claim — product-market fit, retention, LTV:CAC, ARR — reads as developing at best until the evidence ledger backs it. A verified claim lifts a stage more than a logged assumption. Beautiful prose about imaginary interviews can't outscore a real but tersely-described result.

Magnitude checks catch the absurd

Claimed revenue larger than your obtainable market, a SOM bigger than your SAM, a price that single-handedly exceeds the whole market — these get flagged as deal-breakers, not celebrated. An open deal-breaker caps the verdict.

The score is deterministic

No AI decides the headline number. The Advisor (the AI) reads your answers, cites them back, and argues — but the readiness score is computed from your data by rules you can inspect, so it means the same thing every time.

What this looks like in practice

Completeness is not proof. The same field, answered two ways:

Thin — passes the gate

Who has this problem? “Store managers.”

Evidenced — moves the score

Who has this problem? “24 recorded interviews with ops managers at 50–500-SKU groceries; 19 ranked shrink in their top-3 weekly costs.”

…and the Advisor pushes back

“Your buyer is independent cafés, but all six interviews are with enterprise chains — that evidence doesn't support your segment.”

Illustrative — the Advisor cites your own answers, never invented facts.

See it on your own idea.

The fastest way to understand the method is to run a real idea through it.