Ab Test Calculator

This tool evaluates a two-variant experiment (control A vs variant B) from just visitors and conversions per group. It computes each group's conversion rate, the relative lift, a two-proportion z-score and p-value, a 95% confidence interval for the difference, and a significance verdict at the 0.05 threshold, then recommends whether to ship the variant, keep the control, or keep testing.

1. In the Control A card enter Visitors and Conversions. 2. In the Variant B card enter Visitors and Conversions. 3. Click Calculate. Invalid inputs are blocked with toasts: visitors must be > 0, conversions ≥ 0, and conversions cannot exceed visitors. 4. Read conversion rates, lift, z-score, p-value, confidence, the 95% CI, and the Significant / Not Significant badge with its recommendation. Reset clears the form.

rateA = conversionsA / visitorsA; rateB = conversionsB / visitorsB. Standard errors seA = sqrt(rateA × (1 − rateA) / visitorsA), seB analogous. z = (rateB − rateA) / sqrt(seA² + seB²). p-value = 2 × (1 − Φ(|z|)) using an Abramowitz-Stegun approximation of the normal CDF. confidence = (1 − pValue) × 100. lift = (rateB − rateA) / rateA × 100 (0 when rateA is 0). 95% CI for the difference = (rateB − rateA) ± 1.96 × sqrt(seA² + seB²). Significant when pValue < 0.05; recommendation is keepTesting unless significant, then variantWins if lift > 0 else controlWins.