Caused versus credited conversions
Every conversion an advertiser records falls into one of two categories. Some were genuinely caused by the advertising - the customer would not have bought without seeing the ad. Others would have happened regardless and were simply credited to whatever touchpoint sat closest to the sale. Incrementality is the discipline of telling these two apart, because only the caused conversions represent value the advertising actually created.
The distinction sounds academic until you see the gap it opens up. A channel can report thousands of conversions and still be barely incremental if most of those buyers were always going to convert. The reported figure measures correlation - ads were shown, sales occurred - while incrementality measures causation. Standard attribution models distribute credit across touchpoints but cannot, on their own, tell you whether a touchpoint changed the outcome. That is the question incrementality is built to answer.
Why last-click overstates harvesting channels
Last-click attribution awards the entire conversion to the final touchpoint before a sale. By design, this favours channels that intercept buyers at the moment of intent rather than those that create the intent in the first place. Brand search is the classic example: someone who already wants your product searches your name, clicks the paid result, and converts. The click was real, but the demand existed before the ad was ever served. Last-click happily records it as a win for paid search.
The result is a systematic bias toward harvesting channels - brand search, retargeting, and similar bottom-of-funnel activity - while the upper-funnel media that actually generated the demand is under-credited. Teams that optimise against last-click therefore drift budget toward channels that look efficient but create little, and away from channels that create demand but show their payoff later. Incrementality testing exposes this distortion by asking what would have happened if the channel had not run at all.
A worked example: a 9x ROAS that mostly is not real
Consider a retailer running a brand-search campaign at £50,000 a month. The platform reports £450,000 in attributed revenue - a headline 9x ROAS that looks like the best line on the plan. On that number alone, the obvious move is to pour more budget in. But the figure is built on last-click logic, and brand search is a harvesting channel, so it deserves scrutiny before any scaling decision.
To find out, the team runs a geo holdout: brand search is switched off across a representative set of regions while it keeps running elsewhere. Sales in the held-out regions barely move - the customers searching the brand still find it organically and still buy. The experiment shows that only a small fraction of the credited £450,000 was genuinely caused by the paid campaign. The true iROAS is far below the reported 9x - much of that spend was paying to capture demand that would have converted anyway. The reported number was not wrong, it was simply answering a different question from the one that should drive budget.
Why incrementality matters for budget decisions
Media budgets are allocated at the margin, so the decision that matters is never how much revenue a channel was credited with last quarter - it is how much genuinely new revenue the next pound of spend will create. Planning on credited averages quietly rewards channels that harvest existing demand and penalises those that create it, exactly the inversion that last-click bias produces. Incrementality reframes allocation around causal impact, which is the only basis on which budget changes can be trusted.
This is why incrementality sits upstream of efficiency metrics rather than beside them. Once you know the caused outcomes, you can express them as marginal ROAS and compare the return of the next pound across channels - the distinction explored in marginal ROAS vs average ROAS. The forecast-led approach is to estimate incremental return before spend is committed, not to discover it after the fact. You can pressure-test the next increment of budget with the marginal ROAS calculator, and the connection between incrementality and incremental return is unpacked further in what is iROAS.
Related terms
- iROAS - the incremental revenue advertising actually creates, net of baseline demand.
- marginal ROAS - the return on the next pound of spend, not the blended historic average.
- attribution - how conversion credit is assigned across touchpoints, and why it differs from causation.
Frequently asked questions
What is incrementality?
Incrementality measures the additional outcomes that advertising genuinely caused - the conversions that would not have happened without the ad - as distinct from those that would have occurred anyway and were merely credited to it. It separates caused conversions from coincidental ones, giving a truer picture of what a channel actually contributes.
Why does last-click attribution overstate harvesting channels?
Last-click attribution awards full credit to whatever touchpoint immediately precedes a sale, which favours channels that intercept buyers at the moment of purchase. Brand search and retargeting often capture demand that already existed, so they look extraordinarily efficient even when they create little new revenue. Incrementality testing reveals how much of that credited revenue would have arrived anyway.
How do you measure incrementality?
The most rigorous methods are controlled experiments: geo holdouts that switch a channel off in some regions, audience holdouts that withhold ads from a randomised group, and conversion-lift studies. Each compares an exposed group against an unexposed one and treats the difference as the incremental effect. Where live tests are impractical, response and media-mix modelling estimate the baseline a channel would have earned without spend.
How is incrementality related to iROAS?
iROAS, or incremental return on ad spend, is incrementality expressed in monetary terms: it divides only the caused revenue by the spend that caused it. Incrementality is the underlying causal measurement; iROAS turns that measurement into an efficiency ratio you can compare across channels and feed into budget decisions.
Why does incrementality matter for budget decisions?
Budgets are decided at the margin, so the question that matters is how much genuinely new revenue the next pound of spend will create - not how much historic revenue a channel was credited with. Planning on credited averages over-invests in channels that harvest existing demand and under-invests in those that create it. Incrementality reframes allocation around causal impact.