Metrics & Analytics

A/B Testing

A controlled experiment that compares two versions of a product element (A = control, B = variant) to determine which performs better on a defined metric.

What is A/B Testing?

A/B testing (also called split testing) is a randomised controlled experiment that shows two versions of a product element to different user segments simultaneously to determine which version drives better outcomes.


A/B test anatomy

ElementDescription
Control (A)The existing version
Variant (B)The new version being tested
Hypothesis"Changing X will increase Y because Z"
Primary metricThe one metric that determines the winner
Guard metricsMetrics you must not harm (e.g. revenue)
MDEMinimum detectable effect — smallest improvement worth detecting
Sample sizeCalculated from MDE, baseline conversion, and significance level

Statistical significance

A result is significant at 95% confidence when there's only a 5% chance the observed difference occurred by random chance. Don't stop tests early — running until significance is reached after peeking inflates false positive rates.


Common mistakes

  • Too many variants — multivariate tests require much larger samples
  • Stopping early — peeking at results and stopping when significant inflates false positives
  • Wrong primary metric — optimising for clicks while the real goal is revenue
  • No holdout — running a test without a control group

Frequently asked questions

How long should an A/B test run?

Until you've reached the required sample size calculated upfront — not until you see a significant result. For most SaaS products, this is 1–4 weeks. Never stop a test early because it looks like B is winning.

What's the difference between A/B testing and multivariate testing?

A/B testing compares two versions of one element. Multivariate testing compares multiple variations of multiple elements simultaneously. MVT requires much larger sample sizes and is harder to interpret — use it only when you have very high traffic and are optimising multiple elements at once.

Apply A/B Testing to your real product data

PMRead ingests customer feedback, interviews, and Slack threads — and generates PRDs grounded in real evidence.

Try PMRead free →