What is incrementality in marketing?

Incrementality is the difference between the outcomes your marketing actually produced and the outcomes that would have occurred without it. It asks whether an activity changed results or merely sat near results that were already coming.

What is an incrementality test?

A deliberate comparison between a group exposed to a marketing activity and a similar group that was not — via holdouts, geographic splits, staged rollouts, or pausing a channel. The gap between groups estimates what the activity added, though every design only approximates a counterfactual that can never be observed directly.

How is incrementality different from attribution?

Attribution divides credit among recorded touchpoints, so it measures proximity to a conversion. Incrementality asks whether the conversion would exist without the touch. Mature teams often run attribution as continuous monitoring and use occasional incrementality tests to calibrate it.

Can small B2B teams measure incrementality?

Rarely with textbook rigor — samples tend to be small and sales cycles long. But they can ask new customers how they first heard of them, pause a single channel and watch what happens, and write down expected effects before spending, which brings incremental thinking without a formal experiment.

What is incrementality

Guide · Glossary & Definitions · 5 min read · last verified 2026-07-27

Reviewed before publication Editorial board Independent commercial review

In shortIncrementality is the difference between what your marketing produced and what would have happened anyway. How tests approximate the counterfactual, why they are hard in B2B, and how small teams can act on the idea honestly.

Incrementality is the difference between what happened with your marketing and what would have happened anyway. A sale your dashboard attributes to a campaign is not necessarily a sale the campaign caused: the buyer may have arrived through a colleague's recommendation, an old relationship, or a search they would have run regardless, and the campaign simply happened to be the last recorded thing they touched. Incrementality is the discipline of asking the uncomfortable follow-up question — did this activity change the outcome, or merely take credit for it?

The question attribution cannot answer

Attribution systems record touchpoints and distribute credit among them. They answer a narrow question: which of the interactions we could see sat closest to the conversion? That is useful, but it is not the question a budget owner actually cares about. The budget question is counterfactual: would this revenue exist if we had not spent the money?

The gap between those two questions is where marketing reporting most often flatters itself. The classic illustration practitioners point to is advertising on your own brand name: buyers who click those ads frequently already knew the brand and were already on their way. The click gets recorded, credited, and reported, yet many of those buyers would likely have arrived anyway. Attribution sees a touch near a sale and calls it a contribution. Incrementality asks whether removing the touch would have removed the sale — a different question, and usually a harder one. For the rules behind the crediting itself, see What is attribution modeling.

What an incrementality test actually is

An incrementality test is a deliberate comparison between a group exposed to a marketing activity and an otherwise-similar group that was not. The unexposed group stands in for the world where you did nothing; the difference between the groups is your estimate of what the activity actually added.

The common designs vary mainly in how they build that unexposed group:

Holdout tests withhold the activity from a randomly selected slice of the audience and compare outcomes across the split.
Geographic tests run the activity in some regions and not others, treating the quiet regions as the counterfactual.
Staged rollouts introduce the activity to segments at different times, so each later segment briefly serves as a comparison for the earlier ones.
On/off tests pause an always-on channel and watch whether the outcomes it claimed actually dip.

None of these observes the counterfactual directly — no test can, because the world where you did nothing is a world you never get to visit. Every design is an approximation, and the honest ones say so.

Why true incrementality is hard to measure

It would be convenient to report that incrementality testing is straightforward. In business-to-business settings especially, it usually is not, for reasons worth naming plainly.

Samples tend to be small. A market with a few thousand genuine buyers rarely produces enough conversions for a clean split to separate signal from noise. Sales cycles tend to be long, so the effect of an activity may surface quarters after the test window closes. Buying committees complicate exposure: one member may have seen the campaign while the person who signs did not. Groups contaminate each other — recommendations travel through conversations and private channels that ignore your test boundaries, a problem explored in What is dark social. And seasonality, sales pushes, and plain randomness often move outcomes more than the activity being tested does.

The result is that many incrementality estimates in this market are directional rather than precise. That is not a reason to abandon the idea. It is a reason to hold the estimates loosely and to distrust anyone who reports lift with laboratory confidence from field conditions.

Incrementality versus attribution

The two are complements, not rivals, and the mature posture uses each for what it is.

Attribution is continuous, cheap, and always on. It tells you what the recorded paths look like and how they shift, which makes it a reasonable monitoring instrument. Its weakness is structural: it confuses proximity with causation and can only divide credit among the touches it happened to see.

Incrementality is causal in intent but episodic and expensive. You cannot run a holdout on everything at once, and each test occupies a channel for weeks or months. Its natural role is calibration: run attribution day to day, then periodically test whether a channel attribution flatters actually earns its keep. Teams that do this sometimes discover their most-credited channel is mostly harvesting demand created elsewhere — an awkward finding, and precisely the kind worth having before a budget review. How to report marketing to a CFO covers how to present the distinction without drowning the room in method.

A practical posture for teams without a data-science bench

Most small teams will never run a textbook experiment. You can still think incrementally:

Ask every new customer how they first heard of you, in a free-text field, and read the answers. Self-reported memory is imperfect, but it often surfaces influence no dashboard recorded.
When you can tolerate the risk, pause one channel and watch what actually happens to the outcomes it claimed. Interpret with humility — other things move too.
Treat any channel whose reported results look too clean with suspicion, especially channels positioned at the very end of the buying journey.
Write down, before spending, what you expect the spend to change and by when. A prediction made in advance is a small private experiment; a story assembled afterward is not.

Measuring change against a fixed baseline

The deepest habit incrementality teaches generalizes beyond paid campaigns: fix your point of comparison before you act, then re-measure the same way afterward, so that improvement is something you observe rather than something you remember. This is how Magrios treats AI visibility — score your presence in AI answers first, lock that benchmark, do the work, then re-scan against the locked baseline so any movement is measured against a stated starting point rather than against impressions. The same logic underpins How to measure AEO ROI and the broader stance described in What is evidence-backed marketing: claims about what worked deserve the same scrutiny you would apply to anyone else's claims. Incrementality, at bottom, is just that scrutiny applied to your own spending.