Playbook

How we run experiments

The ColdStart Labs methodology: from problem to product in days, structured experimentation instead of launch-and-pray.

01

Problem Selection

Every experiment starts with a problem observed firsthand — not hypothesised from a pitch deck. We look for problems where we have domain context, where the scope is buildable, and where signal will come fast.

Domain knowledgeMarket researchValidation
02

Hypothesis

Frame the bet clearly. What are we building? For whom? What signal tells us it's working? A good hypothesis is falsifiable within weeks, not quarters.

User personaSuccess metricsKill criteria
03

Rapid Build

AI-assisted development takes us from concept to working product in days. We don't build MVPs — we build functional products with real infrastructure. Ship early, not sloppy.

AI-assisted devNext.jsShip fast
04

Launch + Measure

Put the experiment in front of real users. Not a landing page with a waitlist — a working product people can use. Measure actual behaviour, not stated intent.

Real usersAnalyticsFeedback loops
05

Iterate or Kill

Based on signal, either double down or shut it down. No zombie products. Killing an experiment that isn't working is a success — it freed up capacity for the next one.

Decision frameworkPivotResource allocation

The bottom line

Speed is a feature. We'd rather ship six experiments and kill four than spend a year perfecting one. The methodology is designed for learning velocity — not output volume.