What is incrementality in marketing? Most marketing metrics tell you what your ads touched. Incrementality tells you what your ads caused. It is the difference between taking credit and earning it, and it is the single sharpest line between an agency that measures real impact and one that just reads the platform dashboard back to you.
Incrementality, precisely defined
Incrementality is the measure of conversions that happened because of your advertising, and would not have happened without it. You find it by comparing a group exposed to ads against a holdout group that saw nothing, then measuring the lift between them. Unlike attribution, which assigns credit to any conversion an ad merely touched, incrementality isolates the conversions your spend genuinely created.
What is incrementality in marketing, in plain English?
Here is the uncomfortable truth most dashboards hide: a lot of the conversions your ads "get credit for" were going to happen anyway. Someone already searching for your brand, already in the cart, already sold, who happened to click a retargeting ad on the way to checkout. Your platform counts that as a win. Incrementality asks the only question that matters: would that sale have happened if the ad never ran?
If the answer is "yes, probably," then the ad did not cause the conversion. It rode along and took the credit. Multiply that across a whole account and you get the dirty secret of digital advertising: a meaningful share of reported "performance" is the channel claiming sales it did not create.
Incrementality strips that out. You run your ads to one group, you deliberately withhold them from a comparable group, and you measure the gap. That gap, the incremental lift, is the real contribution of your advertising. Everything above the holdout baseline is yours. Everything at or below it was happening regardless.
It is the most honest number in marketing, which is exactly why most agencies never bring it up.
Why incrementality matters
Channels are graded on numbers they get to define. A retargeting campaign reporting a 12x return on ad spend looks incredible until you realize most of those buyers were already going to convert. Strip out the people who would have purchased anyway and that 12x can collapse toward 1x, or worse. The campaign was not driving growth. It was photographing it.
Incrementality matters because it answers the question every budget decision hinges on: if I turned this off, what would I lose? Not what the platform claims I would lose. What I would lose.
That reframes everything:
- Budget allocation. You stop funding channels that harvest demand and start funding channels that create it.
- Scaling decisions. You learn whether the next dollar buys new customers or just re-buys ones you already had.
- Channel fights. When two platforms both claim the same conversion (and they will, because each is measuring in its own walled garden), incrementality is the referee that doesn't have a horse in the race.
For brands running real budgets across programmatic, paid social, and search, this is the difference between a media plan that compounds and one that quietly subsidizes demand you already owned.
How incrementality testing works
Incrementality is measured, not modeled. You cannot infer it from a conversion report. You have to run an experiment with a control group. The core mechanic is always the same: create a group that sees the ads (test) and a comparable group that does not (holdout), then measure the difference in conversions.
The common methods:
Geo lift tests (geo holdouts)
You split markets into matched groups: similar regions, similar baseline performance. Ads run in the test markets and stay dark in the holdout markets. The lift in conversions between them, controlling for market size, is your incremental effect. Geo tests are the workhorse for channels where you cannot cookie or user-match cleanly, including connected TV and streaming audio and a lot of upper-funnel programmatic.
Audience holdouts (ghost ads / PSA tests)
Inside a platform, a randomly selected slice of your target audience is deliberately excluded from the campaign (or shown a neutral placebo ad). You compare conversion rates between the exposed group and the held-out group. This is the cleanest method when the platform supports a true randomized holdout, because the two groups are statistically identical except for one thing: the ad.
Conversion lift studies
Platform-native versions of audience holdouts (Meta Conversion Lift, Google's lift studies, and similar). Useful and convenient, with one honest caveat: the platform is grading its own homework. Treat the result as directional, not gospel, and cross-check against independent geo tests when the stakes are high.
The pattern underneath all three is identical: a baseline that got no advertising, an exposed group that did, and the measured gap between them. No holdout, no incrementality. Anyone who claims to measure "true incremental impact" without withholding ads from someone is selling you a model dressed up as an experiment.
A worked example
Say a retargeting campaign reports 1,000 conversions and a 10x ROAS. On paper, untouchable. So you withhold ads from a randomly selected 10% of the audience and watch what they do. Over the test window, the exposed 90% converts at 6% and the held-out 10% converts at 5.4% on its own, with no ads at all. That 5.4% baseline is demand you already owned. Only the gap is incremental, which means roughly a tenth of those 1,000 "conversions" were caused by the spend. The headline 10x ROAS quietly becomes something closer to 1x once you grade it against the holdout. Same campaign, same dashboard, very different truth. That is the entire reason the test exists.
Incrementality vs attribution
These two get used interchangeably, and they shouldn't. They answer different questions and frequently disagree.
| Attribution | Incrementality | |
|---|---|---|
| Question it answers | Which touchpoints get credit for a conversion? | Which conversions did advertising cause? |
| Method | Assigns credit across the touches that preceded a conversion | Compares an exposed group against a no-ad holdout |
| Counts conversions that would have happened anyway? | Yes | No (that's the entire point) |
| Bias | Over-credits low-funnel and retargeting | Isolates true causal lift |
| Best for | Day-to-day optimization and pacing | Budget, scaling, and channel-value decisions |
Attribution tells you the path. Incrementality tells you the impact. Attribution will happily inform you that retargeting "drove" 30% of conversions; incrementality will tell you how many of those 30% would have converted with no retargeting at all (often, most of them). You need both. Attribution to steer the campaign week to week, incrementality to decide whether the campaign deserves to exist. Run only attribution and you will over-invest in demand harvesting and starve the channels that grow the business. That's not a small error. It's the most expensive mistake in digital media, and it's invisible if you never run a holdout.
For more on how the metrics you're probably watching can mislead, see attribution windows, ROAS (the headline number incrementality exists to correct), and cost per acquisition, which looks very different once you measure it on incremental conversions instead of credited ones.
When you should run an incrementality test
Not every campaign needs one. Tests cost reach (you're deliberately not advertising to someone) and they need enough conversion volume to reach statistical significance. Run one when:
- You're spending real money on a channel and can't confidently say what you'd lose by cutting it.
- Two platforms are both claiming the same conversions and you need a tiebreaker.
- You're about to scale a channel and want to know if the next dollar buys new customers.
- Retargeting or branded search ROAS looks too good to be true. It usually is.
Skip it when volume is too thin to measure a clean lift, or when the cost of the withheld reach outweighs the value of the answer. We'll tell you which situation you're in instead of selling you a study you don't need.
Most agencies won't run a holdout. We will.
Withholding ads from a group to prove your media works takes a confidence most shops don't have, because it risks showing that a channel they're billing for isn't pulling its weight. We measure what's real, even when the real number is inconvenient. If you want a partner who'll tell you which dollars are growing the business and which are just photographing it, that's the entire job.
This kind of measurement discipline runs through how we build and read programmatic campaigns, where the gap between "touched" and "caused" is widest. Want to know what your media is worth? Book 30 minutes or email admin@moonsauceagency.com. No pitch, no pressure, just the real number.
