Incrementality testing
📖 3 min readUpdated 2026-04-19
Meta says it drove 1,000 conversions. But how many would have happened anyway? Incrementality testing measures the real lift your ads create, the conversions that wouldn't have happened without them. It's the closest thing to truth in ad measurement.
The simple test: geo holdout
- Split your market into "ad" and "no-ad" geos of comparable size
- Run ads in the "ad" geo, pause in "no-ad" geo
- Measure revenue in both over 2-4 weeks
- Lift = (ad-geo revenue) - (no-ad geo revenue, adjusted for size)
Why this works
Controls for everything except ad spend. If ad-geo had 20% more revenue than no-ad geo, ads drove that 20% lift.
Platform-supported tests
Meta Lift Study
Meta splits users into test and control. Official holdout. Works if budget is high enough.
Google Conversion Lift
Similar. Google-run holdout.
The truth reveals
Most brands doing incrementality tests find:
- Meta reports 20-50% more conversions than it actually drives
- Retargeting drives far fewer incremental conversions than reported (often 20-40% of what platform claims)
- Branded search largely captures conversions that would happen anyway
- Top-of-funnel channels (YouTube, TikTok) often drive more incremental than reported
How often to test
- Quarterly or semi-annually for major channels
- Before big budget increases
- When MER and platform reports diverge
Cost of testing
You "waste" ad dollars in holdout periods. The learning is worth it, accurate measurement changes budget allocation.
The realpolitik
Many platforms and agencies resist incrementality testing because it reveals they over-claim. Do the tests anyway. The results inform where to invest next dollar.