Lift

Lift answers "did this work, and by how much?" Absolute lift is the raw difference — a control at 4% conversion and a variant at 5% has an absolute lift of +1 percentage point. Relative lift is that difference expressed as a percentage of control: same numbers, +25% relative lift. They sound very different, and people mix them up constantly in launch announcements.

Lift is meaningful only against a baseline and a success threshold. A 20% relative lift on a metric that started at 0.2% conversion is +0.04 percentage points absolute — a fact that matters when you compute revenue impact. The number alone is a marketing claim; the number anchored to absolute units and to pre-committed thresholds is a verdict.

When to use it

Lift is reported on every experiment that produced a measurable result. Make sure the contract specifies whether the success threshold is in absolute points or relative percent — otherwise the verdict turns into a vocabulary argument once the numbers arrive.

What this looks like in practice

The absolute-vs-relative confusion is the single most common source of post-launch overclaiming. "Conversion up 50%!" usually means relative lift on a small base. The same change in absolute terms might be "+0.5 percentage points." Both are true; one is honest. Pre-commit which one your success threshold uses so the verdict is not ambiguous when the team is celebrating or grieving.

Lift can be positive on the variant for the wrong reason. Selection effects, novelty bias, and concurrent campaigns all leak into the number. If a paid push runs during the test window, the lift you measured on a landing-page variant is not landing-page lift — it is "landing-page-plus-extra-traffic" lift. Pre-register confounders in the contract so the verdict accounts for them instead of explaining them away.

Lift compounds in growth-loop products and decays in fashion-driven ones. A signup-flow lift earned in Q1 might still be paying off in Q4. A subject-line lift on a one-off campaign is gone the next day. The verdict in the archive should record whether the lift is durable or ephemeral, so future-you reading the entry knows what to weight it against.

A worked example

Common mistakes

• Reporting relative when absolute is small.+50% relative lift on a 0.4% baseline is +0.2 percentage points. Honest reporting includes both, or someone in the future will assume the wrong one.
• Comparing lift across experiments with different baselines.A 10% lift on a 30% baseline and a 10% lift on a 1% baseline are not the same achievement. Anchor lift to baseline and to operational impact, not to itself.
• Calling lift "incremental" without proof.Lift in an A/B test is causal relative to the variant. Lift on a launched campaign without a control is correlation. Do not confuse the two in the archive — they teach different things.

Related terms