A/B sample splitter
Split sample between A/B/C test variants evenly or by weights
Parameters
Sample split
In SurveyNinja, NPS, CSAT and more are calculated automatically. Create a survey in 5 minutes and get real-time analytics.
Start freeSplit examples
1 Equal 50/50 split
1000 respondents, 2 variants (A/B)
500 per variant A and B — maximum power for comparison.
2 Uneven 70/30
Control (A) larger, test (B) smaller — when the new variant is risky
1000 respondents → 700 for A, 300 for B. Enter shares in the calculator to get the plan.
3 A/B/C — three variants
1500 respondents, 3 variants at 33.3% each
500 per A, B and C. For multivariant test ensure per-variant size is sufficient (see minimum sample calculator).
4 Risky test 90/10
New landing design — show to a small share of traffic first
2000 respondents → 1800 control, 200 test. Reaching significance will take longer.
5 Four variants (A/B/C/D)
Equal shares 25% each. Total sample 4000
1000 per variant. Account for multiple comparison correction when interpreting.
6 60/40 in favor of control
Compromise between power and risk: control slightly larger than test
1000 respondents → 600 for A, 400 for B. Enter 60 and 40 in the calculator.
FAQ about A/B sample splitter
Why split the sample across variants?
When to use even split?
When to use uneven split?
How many variants to test at once?
How to randomize assignment?
What to do after splitting the sample?
How does the splitter relate to the significance calculator?
Can I test more than two variants (A/B/C/D)?
Automate your metrics
In SurveyNinja, NPS, CSAT and more are calculated automatically. Create a survey in 5 minutes and get real-time analytics.
Start free