Experiments

Onboarding-scoped feature flag exposures from experiment.exposed events with surface = 'onboarding'. Each section is a single experiment with its per-variant funnel — exposed → started → completed — and a two-proportion z-test on the completion-rate diff vs the next-best variant.

Window

7d 30d 90d

onboarding_intro_copy_v1

825 exposed · 2026-08-23 → 2026-09-22

control 825 exposed

Started: 794 96.2%
Completed: 29 3.5%

onboarding_flow_config_v1

800 exposed · 2026-08-23 → 2026-09-22

control 800 exposed

Started: 769 96.1%
Completed: 29 3.6%

onboarding_welcome_setup_v1

117 exposed · 2026-09-12 → 2026-09-22

Leader

direct_setup 100% vs 2.6%

z = 5.34 · p < 0.05

direct_setup 1 exposed

Started: 1 100%
Completed: 1 100%

control 116 exposed

Started: 116 100%
Completed: 3 2.6%

Notes

"Started" means the user, after exposure, emitted any onboarding.step_viewed or onboarding.choice_selected. For most exposures this is ≈100% — exposures are usually fired on first paint of an onboarding screen, so the screen view follows immediately. A low started rate is a tell that the flag is being evaluated outside an onboarding session (e.g., a refresh in Settings).
Significance is a two-proportion z-test (|z| ≥ 1.96 → p < 0.05). This is the bar in-app today; switching to a Bayesian credible interval is a Phase 5+ call.
Only experiments with surface = 'onboarding' appear here. Cross-surface experiments live in PostHog Insights — this page intentionally narrows.