← All terms·Digital attribution
Reference library · Digital attribution
Incrementality
Incrementality is the share of conversions caused by an ad exposure, as opposed to conversions that would have happened anyway. It is established by comparing an exposed group against a randomized or matched control group that saw no ad (or a placebo/PSA), not by attributing credit within exposed users alone.
Formula
Worked example
A DTC retailer in Poland runs a holdout test on a YouTube campaign: 400,000 users see the ad (exposed), 400,000 matched users see a PSA instead (control). Exposed CVR = 3.5% (14,000 conversions), control CVR = 2.5% (10,000 conversions). Incremental conversions = 14,000 - 10,000 = 4,000; lift = 4,000 / 10,000 = 40%. Only those 4,000 conversions, not all 14,000, should be credited to the campaign when judging its ROI.
How it is used
Planners use incrementality tests (geo holdouts, PSA/ghost ads, matched-market tests) to validate spend on channels where last-click or multi-touch attribution overstates impact, especially branded search and retargeting. Test cells need enough volume and a long enough window to detect the lift given the campaign's baseline conversion rate, or the result is noise dressed as a finding. The recurring mistake is comparing exposed users to the general unexposed population instead of a randomized or propensity-matched control, which lets self-selection (people already likely to buy click ads) inflate the apparent lift.
The common mistake
Use a randomized or matched control group instead of the general unexposed population, otherwise self-selection bias inflates the measured lift.