Test a variant and get a clear answer

One variable only

An A/B test compares two versions of an ad that differ in just one thing: the image, the hook, or the audience. Change the image and hook together, and if version B wins, you know it won but not why. So you can’t reuse anything.

This discipline is thankless: it slows you down. It’s also the only way to build, test after test, real knowledge of what speaks to your market.

Enough volume to decide

A test needs enough events in each version so the gap isn’t luck. With a handful of results on each side, even a big gap can be pure noise.

If your volume is too low, test on a more frequent event, like clicks or add-to-carts. Just remember you’re measuring a step along the way, not the sale.

Also give the test a full cycle: a whole week, not three days, so weekends and slow days weigh the same on both versions. A test stopped on a Tuesday night often compares two moments, not two creatives.

Decide before you look

Before you launch, write down what you’ll look at, for how long, and what will make you choose. Otherwise, you’ll always find some number that proves your favorite version right.

A simple rule: one main metric, a fixed timeframe, one decision. Keep, replace or relaunch. Nothing else.

What you learn from losing

A losing variant isn’t a failed test. It told you an angle doesn’t work in this market, and that’s worth knowing: you won’t try it again three months from now.

Log every result, winners and losers, somewhere permanent. After a few months, that list is worth more than any generic best-practices guide. It’s about your audience.