FabricFabricExperiments
Concepts

Statistics

Fabric Experiments reports exposure-first aggregate rows and Bayesian summaries.

Fabric Experiments starts with pragmatic, auditable statistics.

Aggregate rows

The aggregate endpoint returns per-variant rows:

  • variantKey
  • exposureCount
  • conversionCount
  • conversionSum
  • failureCount
  • computedAt

Bayesian summaries

The stats package includes Beta-Binomial posterior summaries and normal-approximation credible intervals. This is documented in ADR-0004 and intentionally favors transparent implementation over overclaiming advanced sequential methods.

Deferred

Sequential testing, CUPED, mobile SDKs, Python SDKs, and Go SDKs are deferred beyond M5.1.

Guardrails

SRM, metric decline, and error-rate checks protect experiments while they run.

Audit trail

Experiment changes are recorded as immutable audit events and can be exported or forwarded to agents.

On this page

Aggregate rowsBayesian summariesDeferred