A/B Test Calculator
Sample Size & Duration
Before running an A/B test, you need to know two things: how many visitors each experience needs to collect, and how long that will take. This calculator gives you both — so you know exactly when it's safe to read the results.
New to A/B testing?
Let me walk you through it step by step
What are you measuring?
Best for goals measured by % of visitors who complete an action (click, sign-up, purchase).
Your current numbers
e.g. if 25 out of 1,000 visitors convert, enter 2.5. Find this in your analytics tool.
What's the smallest improvement worth acting on?
10% lift on 2.5% baseline → target rate 2.75%
Your traffic
Not total site traffic — only visitors to the page being tested. Leave traffic % at 100 unless you're running a partial rollout.
Results
Visitors needed per experience
64,215
128,430 total across all experiences
Don't read results until each experience reaches this number
Minimum run time
33 days
at 4,000 visitors/day to that page
Tip: Consider rounding to 35 days to capture full week cycles — day-of-week traffic variation can skew results.
Reminder: Checking results before your planned end date inflates the false positive rate significantly. Trust the plan.
Executive Summary
Executive summary
Run this test for at least 35 days (rounded to full week cycles) before reading results. You need ~64,215 visitors per variant before the result is statistically trustworthy. At your current traffic of 4,000 visitors/day, that's roughly 5 weeks.
Test Reliability
Test reliability
4 / 5 bars — Strong statistical setup
MDE Sensitivity
MDE sensitivity
How duration changes if you adjust how big a lift you need to detect:
| MDE | Sample/variant | Duration |
|---|---|---|
| 5% | 250,917 | 126d |
| 10%← current | 64,215 | 33d |
| 15% | 29,199 | 15d |
| 20% | 16,794 | 9d |
| 25% | 10,984 | 6d |
Activity Note
AT activity note
// EL Calculator — 2026-05-09 // Metric: Conversion Rate | Baseline: 2.5% // MDE: 10% relative lift // Sample: 64,215 per variant (128,430 total) // Duration: ~33 days at 4,000 visitors/day // Confidence: 95% | Power: 80% | Variants: 2