ads-test

★ 7.8 · marketing

ads-test is a Claude Code skill that designs and evaluates paid-media experiments end-to-end, from hypothesis formulation to final decision. It walks through defining the randomization unit, target population, primary metric, guardrail metrics, minimum detectable effect, and stopping rule. Sample size and test duration are calculated from explicitly declared assumptions, with approximations disclosed. The skill accounts for platform constraints across Meta, Google Ads, and LinkedIn — including overlapping experiments, conversion lag, seasonality, and measurement quality issues. Readout includes verifying assignment integrity and data completeness before estimating effect and uncertainty. Output is versioned JSON paired with a plain-language decision; the skill explicitly guards against peeking for favorable results or calling underpowered noise a winner.