revert(marketing): remove landing hero positioning experiment

Merged
SA
3mo ago
+10-174

Summary

Closes out the landing-hero-positioning A/B test (PostHog experiment 386395) and removes all experiment code. Reverts the experiment wiring from #5832 and the flag bootstrap from #5884.

  • HeroSection renders the control copy unconditionally again (no flag read, no platform gating for copy selection)
  • Removed the bootstrap option from the marketing posthog.init (it existed only for instant, sticky hero-variant assignment)
  • Deleted hero-flag-bootstrap.ts (client-side SHA-1 variant assignment mirror)

Result (why we're closing)

After 2 weeks and ~7.8k exposures, download_clicked was flat: control 21.9% vs test 21.7%, 42% chance to win, credible interval [−9.3%, +7.6%] relative — which already excludes the pre-registered +10% MDE, so the test arm can't win. The PostHog experiment is stopped and marked inconclusive; the flag will be cleaned up after this deploys.

https://claude.ai/code/session_016SyQq825A17ZMGGDgyaoPE

Summary by cubic

Remove the landing-hero-positioning A/B test and revert the landing hero to the control copy. Closes PostHog experiment 386395 and simplifies the marketing analytics init.

  • Refactors
    • HeroSection now always renders the control copy; removed flag reads and platform gating.
    • Dropped bootstrap from marketing posthog.init; deleted apps/marketing/src/lib/analytics/hero-flag-bootstrap.ts.

Summary by CodeRabbit

  • Changes
    • Simplified the marketing hero section to use consistent headline and subheadline content.
    • Removed audience-based hero variations and platform-specific positioning.
    • Updated hero animations and sizing to render directly from the standard content configuration.