Experiments

A/B and multivariate tests with persistent assignment and a reliable control group.

Roadmap · Phase 6 Not yet available

Experimentation engine

The statistical core is already built and tested: deterministic hashing assignment, fixed-horizon two-proportion tests, Bonferroni correction and sample-ratio-mismatch detection. Phase 6 connects it to the SDK, exposures and this UI. A winner is never declared before the pre-registered sample size and minimum duration are reached.

TrackPix never shows placeholder numbers. This page will display real results once the module ships — see the roadmap in Reports.

What it will do
  • Experiment setup: page, control, variants, goal, traffic split, audience
  • Persistent randomized assignment — no variant switching mid-session
  • Visitors, exposures, conversions, uplift and confidence intervals per variant
  • Automatic SRM, broken-page and missing-event detection with auto-pause
Requires
  • Event ingestion (Phase 2)
  • Page variants (Phase 5)
  • Experiment edge runtime (Phase 6)