Live
No signup
Conversion Rate Calculator
CR, uplift and statistical significance.
Significance
Statistically significant (95%)
p-value: 0.0323
CR A
2.40%
CR B
3.10%
Lift
+29.2%
How this tool works
The costliest mistake in web optimization is calling a variant winner with too little data. If your A has 2% CR and B has 2.4% with only 200 visits, that's NOT a 20% lift — it's statistical noise.
This calculator applies the 2-proportion z-test, returns real p-value, and tells you whether the difference is statistically significant at 95% confidence (p < 0.05). Same math Google Optimize, Optimizely and VWO use under the hood.
Common pitfalls
- Cutting the test when a variant 'is winning' early (regression to the mean)
- Tests with under 1,000 visits per variant for CR <5% (low statistical power)
- Multiple tests at once without Bonferroni correction (false positives rise)
- Changing the target metric mid-test (cherry picking)
Frequently asked questions
- How many visits do I need minimum?
- Depends on the lift you want to detect. To detect +10% lift on a 3% base CR at 95% confidence: ~14k visits per variant. For +20%: ~3,500. For +50%: ~600.
- What if p-value is 0.06?
- Not significant at 95%. But close. Recommendation: extend the test 2-3 more days. If it stays 0.06+ likely no real difference.
- What's lift exactly?
- (CR_B - CR_A) / CR_A × 100. If A is 2% and B is 2.4%, lift = +20%. Note: lift is reported over base, not absolute percentage points.
- Does it only work for CR?
- Yes — for A/B with binary metrics (converted/didn't). For continuous metrics (revenue per session, time on page) use t-test. For multiple variants (A/B/C/D) use ANOVA.