How We Run 40 Growth Experiments Per Quarter (Without Chaos)

Forty experiments a quarter sounds like a growth-team flex. Done badly, it is forty ways to churn the roadmap. Velocity only compounds when each experiment feeds the next — and that takes an operating system, not a spreadsheet.

Intake to ICE in one week

Every idea lands in a single intake with the metric it moves and the evidence behind it. Weekly, the team scores by impact, confidence, and ease — then commits only what fits the quarter’s primary metric. Anything not tied to that metric waits, no matter how shiny. This one rule kills most of the chaos.

Kill criteria before launch

Every experiment launches with a pre-registered primary metric, a minimum detectable effect, and kill criteria. If the data crosses a kill line, it dies the same day — no sunk-cost debates. The weekly review reads results in fifteen minutes: winners get scaled, losses get documented in a searchable library, and inconclusive tests get a power analysis before anyone blames the sample size.

The result: forty experiments, one compounding memory, zero heroes.