Answer · Conversion Engineering
How Much Traffic Do I Need for A/B Testing?
The short answer
For a typical B2B baseline conversion rate of 3% and a goal of detecting a 20% relative lift at 95% confidence and 80% power, you need about 4,800 sessions per variant — roughly 10,000 sessions per test. Mid-market sites need 30K+ monthly sessions to run primary-conversion A/B tests inside a 4-week seasonality window.
№ 01The longer answer
Sample size depends on three inputs: baseline conversion rate (lower baseline = more sessions needed), minimum detectable effect (smaller lift to detect = more sessions needed), and confidence/power (95%/80% is standard). The math is non-linear; halving the MDE doesn’t double the sample — it quadruples it.
Practical sample-size cheat sheet for 95% CI / 80% power: baseline 2% + 20% MDE = ~7,000 per variant. Baseline 3% + 20% MDE = ~4,800 per variant. Baseline 3% + 10% MDE = ~19,000 per variant. Baseline 5% + 10% MDE = ~11,000 per variant.
Below 30K sessions/month, the workable alternatives are: sequential (Bayesian) testing (cuts required sample 30-50%), pre/post analysis with multi-week baselines (less precise but honest), microconversion-only testing (microconversions happen 10-50x more often, so sample math works at lower volume), or accept that you can’t cleanly test primary conversions and focus engineering on instrumentation + audit-driven changes.
The dishonest alternative many agencies sell: run an under-powered A/B test, stop it when the variant happens to be winning, declare victory. The math says these “wins” are random noise 60-80% of the time. Don’t let an agency report under-powered tests as wins.
№ 02Can I run A/B tests with less traffic if I’m testing big changes?
Yes — bigger MDEs need fewer samples. If you’re testing a redesign with expected 50%+ lift, you can detect it at much lower volume. But 50%+ lifts are rare on already-optimized pages.
№ 03What sample size calculator should I use?
Optimizely’s public calculator and Evan Miller’s “Awesome A/B Tools” (evanmiller.org/ab-testing/sample-size.html) are both reliable. Avoid “CRO experts” who do it on the back of a napkin.
№ 04Should I test microconversions if I can’t test macroconversions?
Yes — microconversions happen 10-50x more often than form submissions, so sample math works at lower traffic. Identify a microconversion that correlates with your macro (e.g., scroll-75% predicts form-submit), then test on that.
Go deeper
Related questions
Three Ways to Start · No Sales Pitch
Want this answered for your business?
$500 audit. 5-day delivery. Refundable on engagement.