mkt-experiment

You run A/B tests by creating experiments from hypotheses, logging variant performance metrics, and applying bootstrap confidence intervals with the Mann-Whitney U test to detect statistically meaningful improvements and select winning variants.

A living marketing playbook is a durable repository that promotes empirically proven winners from your A/B and multivariate experiments, capturing successful rules so future content and campaigns automatically follow validated formats.

Yes, you can test up to 10 variants simultaneously in batch experiments. The engine creates multivariate experiments from your hypotheses, logs performance metrics, and scores each variant to determine statistical winners or discards.

You generate weekly scorecards by configuring the engine to produce cross-channel reports that monitor lead, candidate, and email pacing against your configured targets, tracking experiment progress and alerting you to pacing issues.

Yes, you need YAML-defined skill instructions plus environment configuration including data directory, agent taxonomy, and statistical thresholds. These settings control how the engine applies bootstrap confidence intervals and manages your experiment lifecycle.

Use bootstrap confidence intervals with the Mann-Whitney U test when selecting winners from A/B and multivariate experiments. This combination detects both statistically meaningful and practically useful improvements across marketing channels like content, email, and SEO.