← All terms·Digital attribution
Reference library · Digital attribution
Conversion Lift
Conversion lift is the incremental conversion rate produced by ad exposure, calculated by comparing the conversion rate of users who saw the campaign against a randomized holdout group who did not. It isolates the causal effect of the media from conversions that would have happened anyway, separating true incrementality from simple correlation. A positive lift confirms the campaign drove conversions beyond what baseline demand would have produced.
Formula
Worked example
A campaign reaches 480,000 users in the exposed group, generating 9,600 conversions (2.0% conversion rate). A randomized holdout of 20,000 unexposed users produces 320 conversions (1.6% conversion rate). Lift = (2.0% − 1.6%) / 1.6% x 100 = 25%, and applying the 0.4-point difference to the exposed group's size yields roughly 1,920 incremental conversions attributable to the campaign.
How it is used
Meta and Google run this as a built-in lift test, randomly withholding ads from a control cell within the same auction; independent versions use geo-based holdouts, excluding one region from the media plan, when platform-level testing isn't available. Analysts report lift with a confidence interval, since small holdout samples or low baseline conversion rates make the result noisy. Results get read into budget decisions well before the test reaches statistical power, turning a 25% lift with a wide interval into a false certainty.
The common mistake
The common error is measuring lift against an unreached group, users who simply weren't served the ad due to frequency capping or targeting gaps, instead of a true randomized holdout, which biases the result through selection effects.