Incrementality is the share of attributed conversions that would not have happened without the marketing activity being measured — the causal slice of revenue, stripped of orders that would have landed anyway through organic, brand, or email demand. The operational form is incremental revenue = (revenue with the campaign running) − (revenue that would have occurred anyway). A campaign that drives $100K in attributed revenue while $70K would have landed anyway has $30K of incremental revenue, and a true ROAS roughly a third of what the platform reports.
The second term in that formula is unobservable directly, which is the entire hard problem of measurement. Every practical method is a different attempt to estimate that counterfactual: geo holdouts, audience holdouts via conversion lift studies, ghost ads, marketing mix modeling, seasonality regressions. Incrementality is a measurement discipline, not a number you read off a dashboard.
It earns its place as the concept that exposes why platform-reported ROAS systematically overstates channel value. Attribution credits a touchpoint for every conversion that touched it; incrementality asks whether the touchpoint caused the conversion. Branded search is the canonical case — a brand bidding on its own name reports stellar ROAS, but most of those buyers would have arrived via the organic result. The same logic runs through retargeting and retail-media ads, where the ad sits closest to shoppers already going to buy.
In practice it is the right question but hard to measure cheaply per campaign. Most DTC brands run channel-level lift tests a few times a year and use the measured factor to discount reported ROAS into an incremental ROAS the team plans against. MER is the adjacent measurement: attribution-free, but not incrementality-aware — it can move because organic and brand demand shifted independently of paid. Incrementality closes that loop.