A/B test calculator
Compare two proportions with uncertainty and stated test assumptions.
Com utilitzar A/B test calculator
The A/B test calculator compares two conversion rates from visitor and conversion counts and returns each rate, the difference in percentage points, a confidence interval for that difference and a two-sided p-value. It uses a fixed-horizon pooled two-proportion z test and a Wald interval at a confidence level from 50 to 99.99%.
- Enter the requested values in the settings panel.
- Adjust the basic settings, then open Advanced mode for export and fine-tuning options. Custom settings stay active when the panel is closed.
- Run a/b test calculator, review the output, then use the available copy or download controls.
Què admet aquesta eina
Compare two proportions with uncertainty and stated test assumptions.
- A visitors
- Range: 1 to 100000000
- A conversions
- Range: 0 to 100000000
- B visitors
- Range: 1 to 100000000
- B conversions
- Range: 0 to 100000000
- Confidence (%)
- Range: 50 to 99.99
Límits i processament
Local bounded calculations. Numeric inputs must be finite and within the visible ranges. Lists and tables have operation-specific limits. Models, estimates and supplied rates must be checked against your intended use.
Preguntes freqüents
How is the p-value calculated in this A/B test?
The calculator pools both groups to get a combined conversion rate, computes the standard error sqrt(p(1 − p)(1/nA + 1/nB)), and divides the difference in rates by it to get z. The two-sided p-value is 2 × (1 − Φ(|z|)).
What does the confidence interval for the difference mean?
It is a Wald interval for B's rate minus A's rate, in percentage points, using the unpooled standard error. An interval that excludes 0 points to a real difference at that level, though near the boundary it can disagree slightly with the pooled p-value.
When can A/B test results from this calculator be misleading?
The approximation can break down with small conversion counts, when you check results repeatedly and stop early, or when the same visitor appears in both groups. Decide the sample size in advance and read the result once the test ends. Conversions cannot exceed visitors.
Where is my input processed?
This operation processes its source in your browser. Copy and download are explicit actions; source input is not saved in an account or history.
What are the input limits?
Local bounded calculations. Numeric inputs must be finite and within the visible ranges. Lists and tables have operation-specific limits. Models, estimates and supplied rates must be checked against your intended use.