Brand Lift Studies, Explained: The Math, the Requirements, and the Gap

What a brand lift study actually measures

A brand lift study splits your audience into two groups: people who were exposed to your ad, and a control group who were eligible to see it but didn't. Both groups get a short survey — usually one to five questions — asking things like whether they recall seeing an ad, whether they've heard of the brand, or whether they'd consider buying it. The gap between the two groups' answers is the "lift."

That's the whole mechanism. It's a controlled survey experiment, not a sales report. The platforms report it in a handful of standard metrics: absolute lift (the percentage-point gap between exposed and baseline positive response rates), relative lift (that gap expressed as a percentage of the baseline), lifted users (the estimated number of people whose opinion moved because of the ad), and cost per lifted user (spend divided by lifted users).

Meta, YouTube, and Google all offer some version of this, usually at no extra charge, but only once your campaign clears a minimum spend and reach — and only if a rep or the tool decides you're eligible. Below that line, there's no study, self-serve or managed.

The math behind it — a worked example

Here's how the numbers actually move, using round illustrative figures rather than a real campaign, because the arithmetic is the part every platform help page skips.

Say your exposed group is 500,000 people. The control group, asked the same question, says "yes, I'm aware of this brand" 8% of the time — that's your baseline positive response rate. The exposed group says yes 13% of the time. The absolute lift is 5 percentage points (13% − 8%). The relative lift is 62.5% (5 ÷ 8).

To get lifted users, multiply the absolute lift by the exposed reach: 5% of 500,000 is 25,000 lifted users. If the campaign spent $50,000 to generate that exposure, cost per lifted user is $2.00 ($50,000 ÷ 25,000). That single number — cost per lifted user — is what most brands actually optimize toward once a study closes, and it's worth calculating by hand at least once so you know what the platform dashboard is doing behind the scenes.

What it takes to run one

Every platform gates access behind a minimum spend and a minimum reach, and the setup and wrap-up windows eat real calendar time — plan on two to four weeks end to end, not two days. The stages below are consistent across Meta, YouTube, and Google, even though the thresholds differ.

Two practical notes that don't show up until you're mid-study: eligibility can flip to "not eligible" mid-run if you edit the campaign, and if that happens late, the cleanest fix is usually to stop the study and start a new one rather than trying to patch the old one. Custom questions are allowed on most platforms but usually require extra approval and come with no benchmark norms to compare against — so the standard questions, boring as they are, are what most brands should run first.

What a brand lift study can't tell you

A brand lift study tells you whether people's stated opinion moved. It does not tell you whether they bought anything. Those are genuinely different measurements, and the confusion between them is where most brand-lift budgets go to die.

This is the honest gap: last-click attribution — and survey-based lift — can't prove incrementality, because neither one controls for what would have happened anyway. Only a holdout or a matched control, measured against actual purchases, does that. We run our own display book that way, reconciled in Amazon Marketing Cloud so DSP and sponsored ads stop double-counting each other's conversions. Across 30 advertisers in July 2026, that book produced 57,137 attributed purchases at a blended $5.49 cost per acquisition, 20.1% of them new to the brand — a number a brand lift survey has no mechanism to produce, because it was never built to count purchases in the first place. It was built to count opinions.

Neither number replaces the other. A brand lift study can tell you your recall moved 5 points on the East Coast and flatlined on the West. It can't tell you whether that recall turned into a single sale.

The mistakes we see (including one we've made)

• Treating awareness lift as proof of sales lift. They're correlated sometimes and not others. Report both or neither.
• Launching below the reach threshold. The study runs anyway, collects almost no responses, and reports a result with a margin of error wide enough to mean nothing.
• Comparing exposed vs. eligible-but-unseen without checking frequency. If your control group saw a competitor's ad five times and your exposed group saw yours once, the lift you're measuring isn't the one you think it is.
• Chasing the lift metric the platform pays attention to. We've built display plans that optimized for recall lift because that's what the study rewarded, and let lower-funnel placements starve — the study wasn't wrong, we read too much into it.

None of this is an argument against running brand lift studies. It's an argument for knowing exactly which question they answer before you act on the result.