Validate before you build
Chapter 2 — cheap evidence, real experiments, and killing the idea before it kills your quarter.
Validation is not asking friends "would you use this?" It's running the cheapest possible test that could fail.
What counts as evidence
Ranked roughly by how much you should trust it:
- Money. A pre-sale, a deposit, a paid pilot. Nothing argues with a card charge.
- Committed non-money. A signed LOI, a calendar slot with a target customer, a waitlist with email confirmation.
- Effort. Someone spent 20 minutes configuring a free version, wrote you a detailed problem description, came back twice.
- Opinions. Interviews, surveys, likes. Necessary but weak — opinions are cheap, so treat them as leads to find the evidence above.
Aim to climb this ladder. Ten interviews that end in "cool idea!" are worth less than one stranger paying you $20.
Design experiments, not vibes
An experiment has four parts: a hypothesis stated so it can be wrong, a test, a metric with a target, and an end date. "Build a landing page" isn't an experiment. "Landing page with a $29/mo pricing column; target: ≥3% of 200 targeted visitors join the paid waitlist; one week" is.
Cheap experiments worth running before any code:
- The pricing-page test — traffic to a page with real prices; measure intent, not applause.
- The concierge MVP — deliver the outcome by hand to five people. If doing it manually doesn't scale, that's a finding.
- The 10-interview sprint — with people from the ICP you named, asking about their problem and current spending, never pitching.
- The pre-sale — a checkout page. The strongest signal available in a weekend.
Kill criteria
Decide before the experiment what result kills the idea. The failure mode isn't failed validation — it's endless re-interpretation. "The conversion was low because of the copy" is how a dead idea occupies six months.
What this looks like in practice
Experiments in SaaS Founder OS are exactly this shape — hypothesis, test, metric, target, result, learning — because the stage gates consume them: you don't advance from VALIDATE to BUILD on enthusiasm, you advance on a completed experiment and recorded evidence. If the evidence isn't there, the app will keep telling you — and that's the product working.
Next: Build the small thing.