In one line
Uplift is the performance of a group exposed to advertising minus the baseline — what would have happened anyway even without the ad.
Why it matters
Not every conversion in an exposed group is caused by the ad — some of those people (brand searchers, repeat buyers) would have converted regardless. Uplift isolates the piece that's genuinely incremental.
The standard way to measure it is a holdout test: randomly split users into an exposed group and a holdout group that sees no ads, then compare conversion rates.
Why CPA/ROAS alone isn't enough
CPA and ROAS only tell you what happened among people who saw the ad — they can't tell you whether the ad caused it. If uplift is close to zero, a great-looking CPA can still mean the ad wasn't actually doing much.
Go deeper
See how to run a holdout test to measure uplift in Measuring Incrementality.