Data for Everyone

A/B Testing Calculator

Pick your metric, drop in your numbers, get a confident decision — with the right statistical test for what you're actually measuring.

✓Sample size and duration in one place ✓Free, no signup ✓Runs in your browser — nothing leaves your device
Learn
Learn
Step 1
What are you measuring?
Choose the type that matches your metric — this determines the correct statistical test.
Step 2
Your numbers
Step 3
Test parameters
How long will it take?

Enter your daily eligible traffic to estimate experiment duration.

Reverse mode
What lift can I detect?

The opposite question: given the traffic and time you actually have, this finds the smallest relative lift the test could reliably detect — using the same baseline, α, power and variants set above.

Your result will appear here
Required sample size, estimated test duration, and a plain-English read of what it means.

Two ways to analyse a test: day by day, with the metric of both groups over time — or from the final totals with the classic Z- or t-test.

Test results

·
 

 

The metric in both groups

Difference and confidence interval

Significance by day Hover a day to see its p-value.
p below α that daynot significant
For reference: numbers by group and checks

Table by day
Which file works

Per user — one row per user: date, group, metric value (revenue, or 0/1 for “converted”). The sample dataset looks like this.

date,user_id,group,revenue
2026-09-07,u100000,A,0
2026-09-07,u100006,B,2079

Daily totals — one row per day and group: users, then mean and standard deviation and/or the number of conversions.

date,group,users,mean,sd,conversions
2026-09-07,A,1573,154.5,787.6,114
2026-09-07,B,1613,211.4,872.1,146

Columns are recognised by name (date / дата, group / группа, revenue / value / converted / значение — or users, mean, sd, conversions). Separator: comma or semicolon. There must be exactly two groups; A or control is treated as the control — swap them with the button on the right. Conversion = the share of users with a value above zero.

The sample is synthetic: a true +6% lift in conversion plus a temporary novelty effect in the first week. Every day is computed on all data accumulated up to that day: Welch’s t-test for the mean, the Z-test for conversion — the same tests as in “Final totals”.

Step 1
What are you measuring?
Choose the type that matches your metric — this determines the correct statistical test.
Step 2 · Variant data
A
B
Upload Excel / CSV file One row per user with group and converted (0/1) — or pre-aggregated group, exposed, converted. Columns are auto-detected.
Expected traffic split — used only for the split-health check
Step 3 · Parameters
Revenue projection (optional)