Incrementality
Last updated September 2026
Incrementality measures how much of a conversion would not have happened without a specific ad or channel, isolating true causal lift from conversions that would have occurred anyway. It is typically measured through holdout tests, comparing an exposed group against a matched group that saw no ad, rather than through standard attribution reporting.
Standard attribution answers which touchpoint gets credit for this conversion. Incrementality asks a harder, more useful question: would this conversion have happened at all without the ad. A customer who was already going to buy, brand search included, gets counted as a conversion by attribution either way. Incrementality is the method that separates that guaranteed customer from the one the ad actually created.
01
Why attribution alone overstates performance
Attribution models, even sophisticated multi-touch ones, credit every touchpoint that appears in a converting path, with no way to know whether the conversion would have happened anyway. Branded search is the clearest example: a person who already decided to buy and searches the brand name by name will convert whether or not a branded search ad exists, but attribution credits that ad for the conversion regardless.
02
How incrementality gets measured
The standard method is a holdout test: split a comparable audience into an exposed group that sees the ad and a control group, matched as closely as possible, that does not, then compare conversion rates between the two groups. The difference is the incremental lift, the portion of conversions actually caused by the ad rather than merely counted alongside it.
- 01
Geo-based holdouts
turn a channel off in some markets, compare results against markets where it stayed on.
- 02
Audience-based holdouts
randomly withhold ad exposure from a portion of an otherwise eligible audience.
- 03
Platform-native lift studies
some platforms, including Google and Meta, offer built-in incrementality testing tools.
- 04
Most useful for mature channels, like branded search, where attribution is suspected of inflating true impact.
03
Where this fits for white label clients
Incrementality testing is not run on every channel for every client; it requires enough volume to produce a statistically reliable holdout comparison. Conduit recommends it selectively, usually for channels where attribution data and gut instinct disagree, like a branded search spend that looks essential on paper but might just be capturing demand that existed regardless, the kind of gap that shows up when ROAS looks strong but the underlying KPI vs KPA picture doesn't add up.
