In context
You spend $50K/month on Meta. Click-attributed ROAS says 3.8x. You run a Meta Conversion Lift test with 20% holdout: exposed group converts at 1.8%, control at 1.1%. Incremental ROAS = 2.2x. So the 3.8x was inflated by 60% — real incremental value is lower but still healthy.
Why it matters
- Lift tests reveal true incrementality that click attribution can't.
- Meta runs them natively for spend above ~$30K/month.
- Run every 3-6 months on mature campaigns to recalibrate.
Related terms
Frequently asked questions
How much holdout for a lift test?
10-20% usually. Higher holdout = cleaner signal but more 'lost' revenue.
Can small advertisers run lift tests?
Meta's native requires minimum spend. Third-party tools (Haus, Measured) democratize it somewhat.
Lift test vs MMM — overlap?
Lift tests are surgical (one channel at a time). MMM is broad (all channels together). Use both.