Incrementality

Incrementality is the sales your ads actually caused: the ones that would not have happened at all if the ads had never run.

Attribution answers a generous question, which is "did an ad touch this sale before it happened." Incrementality answers the harsh one, which is "would this sale have happened anyway." Those are very different numbers, because a chunk of the people Meta credits, your loyal customers, people already searching for you, someone who was going to buy today regardless, would have converted with no ad at all. Attribution counts them. Incrementality throws them out and keeps only the sales the advertising genuinely added on top.

The clean way to measure it is a holdout. Meta's conversion lift test takes a random slice of your audience and deliberately never shows them your ads, then compares how much that held-back group buys against everyone who did see them. Whatever the ad-exposed group bought over and above the holdout is the true lift, the part that only exists because the ads ran. It is the closest thing to a real experiment you can run inside the platform, because the control group is people, chosen at random, living without your ads.

Brace yourself, because the incremental number is almost always smaller than the attributed one, sometimes brutally so. Retargeting is the usual shock: campaigns that look like your best performers on attribution often turn out to add very little on top of what was coming anyway, because they spend on people already walking toward the till. That gap is not a bug in the test. It is the test telling you the truth attribution was hiding.

Attribution tells you which sales your ads were near; incrementality tells you which sales your ads actually caused, and the second number is the one your budget should answer to.