Incrementality
Measuring how many conversions were actually caused by an ad vs would have happened anyway.
Incrementality is the share of attributed conversions that were actually caused by the advertising, rather than conversions that would have occurred regardless. Typically measured via hold-out tests or geo-experiments that compare exposed vs unexposed audiences.
Context
Platform-reported conversions are almost never incremental. Meta, Google, and LinkedIn all take credit for conversions that organic demand would have produced. The honest measure is the uplift in conversion rate of the exposed group vs the control.
Incrementality testing is underused because it's uncomfortable — results almost always show that reported ROAS overstates true impact by 30–70%, which is bad news for the marketing team's metrics but good news for the company's profitability.
A geo hold-out test for a DTC brand turned off Meta ads in 20% of US markets for 4 weeks. Attributed ROAS was 3.8x; incremental ROAS (measured by revenue drop in hold-out vs control) was 1.9x — meaning half the 'Meta revenue' was happening anyway.
Incrementality testing requires real scale ($50K+/month on the tested channel) and a 4–8 week window to reach statistical significance. Small accounts can only estimate directionally.
Related terms
Services that apply this
More Analytics terms
CAPI (Conversions API)
Server-side event tracking that replaces pixel data lost to ad blockers and iOS.
Server-Side Tracking
Running analytics and ad tracking from your server instead of the user's browser.
GA4 (Google Analytics 4)
The current version of Google Analytics — event-based, privacy-aware, BigQuery-native.