What incremental lift measures
Incremental lift is the additional result, such as purchases, sign-ups or brand recall, that happened only because people saw an ad. It is measured by splitting an audience at random into an exposed group and a holdout group kept from seeing the ads, then comparing the two. The difference between the two groups is the lift; everything else would have happened anyway.
Platform attribution counts conversions that followed an ad, whether or not the ad caused them. Lift measurement asks the causal question. Platforms offer two common versions: conversion lift studies, which measure sales or actions, and brand lift studies, which survey exposed and unexposed people about recall or intent. MER is a blended alternative when a clean holdout is not possible.
How brands running creator ads use it
Lift tests answer questions attribution cannot: does whitelisted creator content add sales beyond what the brand account already drives, or do retargeting ads mostly reach people who were going to buy? They need enough budget and conversions to produce a clear result, so most brands run them a few times a year on their largest bets.
Geographic holdouts are the do-it-yourself version. Withhold a campaign in a set of comparable regions, keep it running elsewhere and compare sales. It is less precise than a randomised platform study, but it works across channels, including creator posts and whitelisting.
Worked example
A snack brand runs a four-week conversion lift study on its TikTok creator ads. The exposed group produces 3,000 purchases; an equal-sized holdout produces 2,400 without seeing any ads. The incremental purchases are 600, a lift of 25% over the holdout baseline.
TikTok's own attribution had credited the campaign with 1,800 purchases. Spend divided by 600 gives the true incremental cost per purchase, three times what the dashboard showed. From then on the team uses the incremental figure, not the attributed one, when deciding how much to spend behind creator ads.
Common mistakes
Running a test that is too small is the most common error. With few conversions, random noise swamps the difference and the result is inconclusive, which reads as no lift when the test simply lacked power. Check the platform's minimum requirements before launching.
The other is contaminating the holdout. If the holdout group still sees the same creator content through a creator's own organic post or a parallel campaign, the gap shrinks and the ads look weaker than they are. Plan the test around everything else that is running, and size budgets with the ROI calculator.
How it's used
"The lift study says 600 extra orders, not the 1,800 TikTok reports, so budget off the incremental CPA."
"Let's hold out the Midwest for four weeks and see what sales do without the creator ads."