Uplift is the improvement in a result that can be attributed to a specific change, measured against what would have happened without it. If a new landing page converts 4.0% of visitors and the old one converted 3.2% over the same test, the uplift is 0.8 percentage points, or 25% in relative terms.
How uplift works
Uplift always needs a comparison. In an A/B test, visitors are randomly split between a control and a variant, and uplift is the difference between them. In a conversion lift study run by Meta or Google, a randomly chosen group is held back from seeing the ads, and uplift is the extra conversions among people who could see them. In a regional test, areas with and without a change are compared.
There are two ways to express it, and mixing them up causes confusion:
- Absolute uplift is the raw difference, such as 0.8 percentage points of conversion rate or £3 of average order value.
- Relative uplift is the difference as a share of the starting point, such as 25%. Relative figures flatter small baselines: moving from 0.4% to 0.6% is a 50% relative uplift but only 0.2 points.
Every uplift figure comes with uncertainty. A test with a few hundred visitors per side might show a 25% uplift while the true effect could plausibly be anywhere from slightly negative to strongly positive. Statistical significance and a confidence interval tell you how much of the measured uplift to believe.
Why it matters
Uplift is the honest test of whether marketing works, because it asks what changed because of your action rather than what happened alongside it. A retargeting campaign may report hundreds of sales, but if most of those people would have bought anyway, its uplift is small. That is the same idea as incrementality, applied to a specific change.
For smaller UK businesses, the practical limit is traffic. Many sites do not get enough visitors to detect a modest uplift in a sensible time, so it pays to test bigger changes, such as a new offer, a shorter form or a rewritten page, rather than button colours.
Common mistakes
- Quoting relative uplift without the baseline, so a tiny change sounds dramatic.
- Stopping a test the moment it shows a big uplift; early results swing widely and usually settle lower.
- Comparing a before period with an after period instead of running control and variant together, so seasonality, a bank holiday or a competitor’s sale gets counted as uplift.
- Treating uplift in clicks or add-to-baskets as uplift in revenue.
- Assuming the uplift will last; some of it can fade once the change is no longer new to returning visitors.
How to act on it
Before a test, decide what size of uplift would be worth acting on and which metric it applies to, ideally leads or revenue. Work out whether your traffic can detect it; if not, test a bolder change or run the test for longer. When it ends, report absolute and relative uplift together, with the range of uncertainty. My A/B test significance calculator does the arithmetic.
For ad spend, ask the platform about a lift study if your budget qualifies, or hold back a region or audience yourself. Designing tests that measure genuine uplift rather than activity is part of how I run performance marketing.