Incrementality testing measures how many sales your ads caused that would not have happened without them. You withhold ads from a random or geographic holdout group, compare its results with the group that saw ads, and treat the difference as the incremental result. Platform attribution cannot answer this question on its own.
Why attribution is not enough
Attribution gives credit to ads a customer touched before buying, but it cannot tell whether they would have bought anyway. The gap is widest in branded search, retargeting and repeat or discount-driven customers, where reported ROAS can look excellent while the money the ads truly added is far smaller.
How a test works
Split the audience, or a set of markets, at random into an exposed group and a holdout group. Show ads to the exposed group only, then compare the same outcome in both over the same period. Whatever the exposed group did beyond the holdout rate is incremental.
A worked example
The numbers are invented to show the arithmetic. A campaign spends $30,000 and the platform reports 1,500 purchases, or $20 each.
| Group | People | Purchases | Conversion rate |
|---|---|---|---|
| Exposed to ads | 90,000 | 1,800 | 2.00% |
| Holdout (no ads) | 10,000 | 160 | 1.60% |
At the holdout rate, the exposed group would have made 1,440 purchases anyway, so the ads added 360, a lift of 25%. The real cost per incremental purchase is $83, not $20.
Designing a test that holds up
The main designs are platform conversion lift studies, geo experiments, on/off tests and marketing mix modelling. Which one fits depends on budget, channels and conversion volume. Most failed tests fail on design: weak tracking, too few conversions, or changes made mid-test. That is where help pays for itself, and it is what our testing and incrementality measurement service covers.
Further reading
Frequently asked questions
Is incrementality testing the same as A/B testing?
No. In an A/B test both groups still see ads. An incrementality test includes a group that sees none, which is what lets you measure what the ads caused.
What is incremental ROAS?
Revenue caused by your ads, as measured by a test, divided by ad spend. It is usually lower than platform-reported ROAS because it removes sales that would have happened anyway.