Running two versions of a page, an ad or an email at the same time, splitting traffic between them, and keeping whichever produces more of the outcome you actually want.
The value of an A/B test is that it removes opinion from a decision. Without one you are comparing this month to last month, and a hundred other things changed in between — the weather, a competitor's promotion, a long weekend.
The catch for most small businesses is volume. A test needs enough conversions on each side for the difference to mean anything. At twenty enquiries a month, a test that looks like a 30% improvement is usually noise, and acting on it makes things worse rather than better.
| Visitors per version | 500 | 5,000 |
|---|---|---|
| Conversions A | 12 (2.4%) | 120 (2.4%) |
| Conversions B | 17 (3.4%) | 170 (3.4%) |
| Apparent lift | +42% | +42% |
| Could this be chance? | Easily | Very unlikely |
Identical percentages. At 500 visitors the difference is five conversions, which is a normal week-to-week swing. At 5,000 it is a finding.
This comes up most often in Conversion Rate Optimisation. Finding and fixing the points where visitors leave — usually cheaper than buying more visitors.
A free consult is twenty minutes and there is no pitch deck at the end of it.