A/B testing
Definition
A/B testing compares two versions of one element, such as a headline, an offer or a page, by showing each to a randomly split share of traffic at the same time and measuring which produces more conversions. It needs enough traffic and one change at a time.
Decide the metric, the split and the stopping rule before the test starts, and run it for whole weeks so every weekday gets equal weight. The federal government's weekend is Saturday and Sunday, with a half day on Friday, while private companies set their own working week, so check how your customers' week actually runs. A test that straddles an event changes the audience underneath it: Ramadan is expected to begin on 8 February 2027, subject to moon sighting, and the White Friday sales arrive in late November.
Most small sites can't run a clean test. A few dozen conversions a month can't separate a real lift from luck, so on low traffic use customer calls, session recordings and dated before-and-after comparisons, and save formal tests for pages with volume. In ad platforms, two ads in one ad set aren't an A/B test, because delivery tilts toward the early favourite. Use the platform's experiment or split-test tool so audiences don't overlap.
Definitions are free.So is the cost review.
A short call, then a written review inside 72 hours: what your marketing costs, what it produces, and what is worth fixing first.