ab split test calculator

This calculator uses a frequentist two-proportion z-test (two-tailed) for statistical significance.

What this A/B split test calculator tells you

An A/B split test compares two versions of a page, button, ad, or checkout flow to see which one converts better. This calculator focuses on conversion outcomes and helps you answer one practical question: is the observed lift real, or could it just be random noise?

You enter visitors and conversions for version A and version B. The tool returns conversion rates, absolute and relative lift, z-score, p-value, and a confidence interval for the difference in conversion rates. That gives you both a direction and a level of certainty.

How to use it correctly

  • Run both variants at the same time to avoid seasonality bias.
  • Track the same conversion event for both variants.
  • Avoid stopping the test too early after a few hours of data.
  • Use enough traffic before drawing conclusions.
  • Decide your confidence threshold before launching the experiment.

Key metrics explained

1) Conversion rate

Conversion rate is conversions divided by visitors. If version A has 120 conversions from 1,000 visitors, A converts at 12%. This is the base measurement.

2) Absolute lift vs relative lift

Absolute lift is the direct difference between rates (for example, +2.5 percentage points). Relative lift compares against control (for example, +20.8% relative improvement). Teams often discuss relative lift, but both values matter.

3) P-value and significance

The p-value estimates how likely your observed difference would appear if there were actually no real difference. A lower p-value means stronger evidence against “no effect.” If your p-value is below your alpha threshold (for example, 0.05 at 95% confidence), the result is considered statistically significant.

4) Confidence interval

The confidence interval gives a plausible range for the true conversion-rate difference (B minus A). If the entire interval is above zero, B likely beats A. If it crosses zero, the winner is uncertain.

Practical interpretation framework

After running the calculator, evaluate results with this sequence:

  • Direction: Is B above or below A?
  • Certainty: Is the p-value below your chosen threshold?
  • Magnitude: Is the lift large enough to matter to revenue?
  • Operational impact: Is implementation easy and low-risk?

A statistically significant result with tiny business impact might not be worth shipping. Conversely, a meaningful lift with weak significance may justify collecting more data.

Common A/B testing mistakes to avoid

  • Peeking bias: checking significance repeatedly and stopping the moment it “wins.”
  • Traffic mismatch: uneven randomization or tracking bugs that skew assignment.
  • Multiple changes at once: changing headline, image, and offer simultaneously obscures causality.
  • Ignoring segmentation: a global win may hide losses in key channels or devices.
  • Novelty effects: short-term spikes that fade after launch.

When you should run a longer test

Extend test duration if:

  • Your confidence interval still crosses zero.
  • Daily traffic varies heavily by weekday/weekend behavior.
  • Conversion is rare and you need more observations.
  • You suspect delayed conversions (long consideration windows).

Final takeaway

A/B testing is a decision framework, not just a math exercise. Use this calculator to quantify confidence, then combine statistical evidence with business context. Over time, a disciplined experimentation process compounds into major gains in conversion rate optimization, user experience, and revenue.

🔗 Related Calculators