Intuition isn't a testing strategy
You built a new onboarding tooltip. It looks great. But is it actually better than what you had before? Without A/B testing, you're shipping changes based on intuition, not evidence.
How A/B Testing works
Create variants
Duplicate any existing flow directly from the Flows page, modify the variant — different copy, different trigger, different step count — and set your traffic split percentage.
Deterministic assignment
Users are assigned to variants by a hash of their session ID. The same user always sees the same variant — no flickering, no confusion, no variant-switching between page loads.
Statistical significance
A two-proportion z-test runs continuously as data accumulates. When sample size is too small, you see "need more data." When results are statistically significant, you see the winner — not noise.
What you get
- Split any flow into A/B variants directly from the Flows page
- Adjustable traffic split percentage — 50/50 default, configurable per test
- Deterministic variant assignment via session ID hash — consistent user experience
- Conversion rate comparison across variants with live updating counters
- Two-proportion z-test for statistical significance — real math, not gut feel
- "Need more data" indicator when sample size is too small to conclude
- Declare winner and archive loser — winner takes 100% of traffic going forward
Pricing
A/B Testing for Flows is available on Deploy ($249/mo) and above. Deploy includes the full no-code flow builder and unlimited flows. Scale ($449/mo) adds Slack alerts for statistically significant results plus AI replay diagnosis on every losing variant.