Incrementality is the change in outcomes caused by an activity compared with what would have happened without it. A randomized holdout can estimate that counterfactual for an eligible population.
If equally sized assigned groups produce 120 purchases with a campaign and 100 without it, estimated lift is 20 purchases for the treated-size group. The estimate has uncertainty; credited conversions may be much larger.
Hypothetical equal-size randomized groups and follow-up. The campaign group has 120 purchasers versus 100 in holdout. Estimated lift is 20 purchasers for a treated-size group, or two percentage points. Its baseline is inferred from holdout; individual counterfactuals and statistical significance are not established.