experiment-design

To turn a hypothesis into a verifiable experiment design, you input the recommended hypothesis and OKRs to generate a structured manifest.yaml. This manifest formalizes the experiment by defining specific variants, measurable metrics, and gate conditions required for validation.

AB-test planning requires defining measurable metrics, variants, and design parameters. The generated manifest calculates specific plan parameters including experiment duration in days, minimum sample size, and statistical power to ensure the test is properly sized.

Yes, you can use feature flags and fake door tests for hypothesis validation. The experiment design supports multiple types including ab_test, feature_flag, concierge, wizard_of_oz, and fake_door designs to apply to various validation scenarios.

A manifest.yaml file is used in experiment design to provide a ready-to-run structured output for tracking and governance across teams. It includes the experiment_id, hypothesis, variants, metrics, design, gates, and evidence_refs needed for measurable validation.

To calculate duration and sample size for an experiment, the design process applies defined variants and metrics to compute plan parameters. It outputs the estimated duration_days, min_sample, and power needed to achieve measurable validation within the manifest.