Conversion Lift

Conversion Lift is Meta's own holdout-based incrementality test, where a randomly chosen control group is deliberately never shown your ads so the extra conversions from the exposed group reveal what the ads actually caused.

You have stared at ROAS long enough to know it answers the wrong question. It tells you how many sales the platform is willing to claim, not how many of those sales would have happened anyway. Conversion Lift is Meta's attempt to answer the real question. It splits your audience at random, holds one slice back and never shows them the ads, then compares the two groups. The difference in conversions between the exposed group and the untouched control is the lift, and that is as close to true cause as you get inside Meta.

You reach for it when a number feels too good to trust. If a retargeting campaign reports a huge return, a lift test can show you that most of those buyers sat in the control group and bought at nearly the same rate, which means the ads added little. Run the same test on a cold prospecting campaign and you often find the opposite, real incremental sales that click attribution was underselling. It quietly reorders your sense of what is actually working.

The honest catch is that access is limited and the test only tells the truth at real scale. Conversion Lift has usually needed a Meta rep or partner to set up, and it takes meaningful spend and conversion volume to reach a trustworthy result, so check the current requirements in your own account rather than assuming it is a self-serve button. A small account with thin conversion numbers gets a noisy result that looks like a verdict but is not.

Attribution tells you who to thank; lift tells you who actually earned it.