What is incrementality?
Incrementality is the share of sales that would not have happened without the marketing that ran, as opposed to sales the reporting merely took credit for.
Formula
Incremental revenue = revenue with the marketing running − revenue that would have happened anyway
Worked example, not a result
A holdout test read properly
Say a store splits its market into two comparable groups. In one it keeps advertising as normal, in the other it pauses a single campaign type for a defined period.
The group with advertising takes 16,000 in revenue. The group without takes 13,450. The difference of 2,550 is the incremental revenue, so that campaign type produced far less new business than its own report claimed, because most of the revenue arrived either way.
That result is common with campaigns aimed at people who already know the store. Someone who searched for your brand name, or who viewed a product yesterday, was often going to come back regardless, and the ad simply gets in front of a purchase that was already happening.
A test like this is blunt and needs a long enough period and groups that genuinely behave alike, otherwise seasonality does the work rather than the advertising. The figures above are an example.
Why it matters
Incrementality answers the only question that really matters about ad spend, which is whether it made the business bigger. Attribution answers a different question about which touchpoint deserves credit, and a campaign can score well on that while adding almost nothing.
This matters most on the visitors lever, where the temptation is always to keep funding the campaigns with the best reported numbers. Those are usually the ones closest to the purchase, and they are also the ones most likely to be taking credit for demand that already existed.
For a store owner the useful discipline is occasional and deliberate. Pause a campaign type in a defined way, watch total revenue rather than platform revenue, and find out what actually changes. It is uncomfortable and it is the only way to know whether a budget line is buying customers or buying credit for customers you had.
Common mistakes
- Testing for too short a period. Buying cycles run longer than a week in most categories, and a short test measures noise.
- Comparing groups that are not alike. Regions with different weather, different competition or different delivery times will diverge whatever the advertising does.
- Reading platform reports during the test as evidence. The point of the test is that platform reporting is the thing being questioned.
- Pausing everything at once. Then you learn only that marketing works, which you already knew, rather than which part of it was carrying the result.
Related terms and answers
Your plan is free.
A person reads it. You keep the plan. No logins, no call first.