Incrementality is the part of a result that happened because of the advertising, and wouldn’t have happened without it. It’s measured by comparing what happened with the ads against what would have happened without them, usually with a test group and a control group. Platform attribution can’t show incrementality on its own, because it credits conversions that occurred after an ad, not conversions that occurred because of it.
What incrementality means in advertising
Incrementality answers one question: how much of this result would have been lost if we hadn’t run the campaign?
A customer who was already going to buy, and saw an ad on the way, counts as a conversion in attribution. In incrementality, that sale doesn’t count, because it would have happened anyway. The difference matters most for brands with strong existing demand, for retargeting and for peak periods, where many conversions would arrive regardless of spend.
Attribution vs incrementality
Attribution distributes credit for conversions that happened. Incrementality estimates how many conversions the advertising actually caused.
| Question it answers | Attribution Which touchpoints came before this conversion? | Incrementality Would this conversion have happened without the ads? |
| How it’s calculated | Rules or models over observed paths | Comparison against a control group or baseline |
| Typical bias | Over-credits channels close to the purchase | Depends on test design and sample |
| Best used for | Optimizing inside channels | Deciding how much each channel deserves |
Both have a role. The problem starts when attribution is used to answer the question only incrementality can answer.
How incrementality is measured
Incrementality is measured by comparing an exposed group with a comparable group that didn’t see the ads, or by modeling what would have happened without them.
The most common approaches are holdout tests, where part of the audience is deliberately not shown ads, geo tests, where regions are compared against each other, and marketing mix modeling, which estimates each channel’s contribution from historical data. Each has trade-offs in cost, speed and precision. Results should be described as estimates, with their method and period stated, not as exact measurements.
Why incrementality changes budget decisions
Incrementality changes budget decisions because the channels that look best in attribution aren’t always the ones that add the most.
Channels that capture demand close to the purchase tend to look very efficient in attribution, while channels that create demand earlier tend to look weaker. Moving budget based on attribution alone can therefore cut what actually drives growth. That’s why the gap between platform-reported ROAS and your real KPI matters so much, and why incremental evidence is the strongest argument when you need to defend your media budget to the CFO.
Where Mainkore stands on incrementality
Our position at Mainkore is that the result of media buying should be verifiable by someone other than the party that decided the spend.
That’s why measurement methods run by independent third parties, such as marketing mix modeling, are the most useful way to evaluate any media buying system, including ours. How the KPI Mainkore commits to is defined and measured is explained in how a guaranteed KPI is measured.
FAQ
The part of a result caused by the advertising, which wouldn’t have happened without it. It’s estimated by comparing exposed and unexposed groups or by modeling a baseline.
Attribution distributes credit among touchpoints that came before a conversion. Incrementality estimates whether the conversion would have happened without the ads.
With holdout tests, geo tests or marketing mix modeling. Each involves trade-offs in cost, speed and precision, and results should be reported as estimates with method and period.
Measurement is one of the questions behind the eight criteria in our media buying models guide. Want to talk about how to verify what your media is adding? Talk to us.
Mainkore. Results you can verify without asking us.
