Attribution and incrementality are not the same thing, and treating them as the same is the single most expensive measurement mistake in retail marketing. Attribution is a story about which click came before checkout. Incrementality is the answer to the question that actually matters: would the customer have bought anyway?
The shortest possible definition
Incremental revenue is revenue that exists because of an ad — revenue that would not have been earned without it. Non-incremental revenue is revenue an ad got credit for but would have closed anyway, often through a returning customer, organic search, or a branded keyword.
A 6x ROAS branded-search campaign can have an incrementality of 5%. Most of the revenue would have arrived through organic typing of the brand name. The campaign earned 5% of the credit it reported.
Three ways to actually measure it
There are three credible methods. Each one has a place; none of them is a black box.
- Geo holdouts — pause spend in a matched set of geographies for a fixed window and compare to control geos. Cleanest read for upper-funnel channels.
- Conversion lift studies (platform-native) — useful as a directional check, but the platform is grading its own homework. Treat as one input, not the answer.
- Ghost bid tests — run identical campaigns at a meaningfully reduced bid in a held-out segment and observe what happens to revenue. Fast read on bid-driven incrementality.
Where Campaign-IQ fits
Campaign-IQ does not replace incrementality testing — it makes the result actionable. POAS already discounts revenue you should not be paying for (refunds, returns, fee drag). Layering an incrementality coefficient on top tells you what fraction of the remaining POAS is causal, not coincidental.
Most operators do not have the budget or the calendar for monthly geo holdouts. The pragmatic path is one quarterly incrementality calibration per major channel, plugged back into the agentic recommendations so the system stops over-crediting non-incremental spend.
- Attribution tells you the order of clicks; incrementality tells you whether the order would have happened anyway.
- Branded search, retargeting, and returner audiences are the highest-volume sources of non-incremental credit. Test them first.
- You do not need to test continuously. One credible calibration per quarter, fed back into POAS, beats six platform-graded reports.