A/B testing

Also called Split testing, Multivariate testing

Showing two versions to different people to find out which performs better. Needs more traffic than most small businesses have.

What it means

An A/B test splits an audience at random between two versions of something — a page, a subject line, an advert — and compares a chosen outcome.

Random assignment is what makes it evidence rather than a comparison of two different weeks.

Why it matters

It replaces opinion with measurement on questions that are otherwise argued about indefinitely.

The honest caveat is volume. At low traffic the difference you observe is usually noise, and acting on it is worse than not testing, because it comes with false confidence.

What it looks like in practice

Use a sample size calculator before you start. If the required numbers are far beyond your traffic, do not run the test — make the change on judgement and measure the trend instead.

Test one meaningful change at a time. Button colours rarely matter; offers, headlines and how much you ask for usually do.

Decide the success metric and the duration in advance, and run at least a full week to cover weekly patterns.

What to watch out for

Stopping as soon as a result looks good. Early leads reverse regularly, and stopping on a peak is how a difference gets manufactured.

Testing during an unusual period — a sale, a holiday, a PR mention.

Where to get proper advice

Free sample size calculators are widely available. Most email and advertising platforms have testing built in.