Growth experiment
Definition
A growth experiment is a structured, time-boxed test designed to validate or disprove a specific hypothesis about what will move a product or business metric.
Growth experiments are distinct from feature development. A feature is built when the team believes it's the right thing to build. An experiment is run when the team wants to find out whether something is the right thing to build.
A well-formed growth experiment has four parts: hypothesis (if we do X, we expect Y because Z), metric (the one number that will move if the hypothesis is true), minimum viable test (the cheapest way to generate signal), and decision criteria (what result counts as confirmation vs. refutation).
The cheapest tests come first. If you can test a hypothesis with a five-question user survey, do that before building a prototype. If a prototype is needed, build the scrappiest version. Only run a full A/B test when you need statistical confidence at scale.
Common failure modes: running experiments without a pre-registered hypothesis (you'll find a confirming signal in the noise), optimizing local metrics that don't connect to the north star, and running too many experiments simultaneously so results are uninterpretable.
Experiment design sessions are whiteboard-heavy. Teams sketch the hypothesis tree, identify the metric, debate the minimum viable test, and define success criteria. BoardSnap captures those sessions so the experiment's rationale is documented before the first test runs.
Examples
- Testing whether a new onboarding tooltip increases activation by 15% in a two-week cohort test
- Sending a different subject line to half a re-engagement email list to test open rate
- Building a five-screen prototype to test a new navigation concept before any engineering work
- Running a five-user moderated test before committing a feature to the roadmap
Snap a growth experiment. Ship its actions.
BoardSnap turns any whiteboard — including this one — into a summary and action plan.