Growth Experiment Playbook Template

It depends on traffic. A team with 100K monthly visitors can typically run 4-6 A/B tests per month with sufficient statistical power. A team with 10K visitors may only support 1-2 tests with meaningful sample sizes. The constraint is almost always traffic, not ideas. Focus on fewer, higher-quality experiments rather than rushing many underpowered tests.

Industry benchmarks suggest a 15-30% win rate for well-run growth teams. If your win rate is above 50%, you are probably running safe, incremental tests and missing larger opportunities. If it is below 10%, your hypothesis generation needs work. Track win rate quarterly to calibrate your experiment quality.

[RICE](/frameworks/rice-framework) is better for teams that need rigor and cross-functional alignment because it forces explicit estimates for Reach, Impact, Confidence, and Effort. [ICE is simpler](/compare/rice-vs-ice-vs-moscow) and works well for small teams moving fast. Pick one and use it consistently. The scoring matters less than the discipline of scoring at all.

Document them as inconclusive, not as failures. An inconclusive result means you need more traffic, a bigger effect size, or a different approach. Record what you learned, note the sample size achieved, and decide whether to extend the test, redesign the variant, or move on. Never call a test "negative" just because it did not reach significance.

Start with one experiment per sprint. Make the process visible: share hypotheses in standup, review results in retro. Celebrate learnings from failed experiments as loudly as wins. After 2-3 months of consistent execution, the team will have enough data to justify expanding the program. ---