A brand lift study is the measurement brand teams trust most and understand least. It carries the authority of a controlled experiment, it produces a number a CMO can put in a deck, and it is the closest thing digital advertising has to a common currency for brand effect. It is also expensive, retrospective by design, and structurally unable to answer the question most teams are actually asking. All of that is true at once. Here is what a lift study measures, what it costs, and where it stops being the right tool.
What a brand lift study actually measures
The mechanism is a randomized experiment. The platform splits your audience into two groups: one is eligible to see the campaign, one is held out. Both groups are then served a single survey question. The difference in positive responses between them is the lift — the part of the change that the advertising, and not the market, produced.
On Google you pick up to three from a fixed list: ad recall, awareness, consideration, favorability, purchase intent, brand association, or a custom question. Meta runs a comparable poll-based design. The metrics sit on a ladder, and they do not move at the same speed:
- Ad recall — did they remember seeing it? The most responsive metric, and the one that moves first.
- Awareness and association — do they know the brand, and connect it to the benefit you claimed?
- Consideration and favorability — attitude shifts. Slower, noisier, harder to move in one flight.
- Purchase intent — the metric everyone asks for and almost nobody clears. It needs the largest effect to register above noise.
What it costs to get in the door
The minimums are the part that surprises people. Google publishes required media spend by country tier, and the money has to be spent within a 10-day window after the study starts:
- US and UK — $10,000 for one question, $20,000 for two, $60,000 for three.
- Canada, Germany, Japan — $15,000, $30,000 and $60,000 respectively.
- India, Brazil, Mexico — $5,000, $10,000 and $20,000.
Eligibility is narrower than most teams expect too: Google Brand Lift runs on video, audio or Demand Gen campaigns. Meta sets its own per-country minimums rather than one global figure, on the same logic — and that logic matters. These are not billing thresholds, they are statistical power thresholds. The platform needs enough exposed users and completed surveys to separate a real lift from sampling noise. Underfund a study and you do not get a smaller answer; you get an inconclusive one, after the money is gone.
The four things it cannot tell you
None of this makes lift studies bad. It makes them a specific instrument with a specific blind spot.
- It is retrospective by construction. A lift study requires live media, which means you learn whether the creative worked only after you have committed the budget that would have paid for finding out.
- It measures the campaign, not the asset. With six assets in rotation, the result is a blended read. Isolating one creative means funding a study per creative — at $10,000 a question in the US and UK, that arithmetic ends the conversation quickly.
- It reports, it does not diagnose. A flat purchase-intent number tells you the ad did not move people. It will not tell you the hook lost them in two seconds, or that they recalled a competitor.
- Small brands rarely clear the bar. Real effects below a couple of percentage points routinely fail significance, and “no measurable lift” gets read internally as “the campaign failed” when it often means “the study was underpowered.”
Where pre-testing fits
The two methods answer different questions in a fixed order. Pre-testing decides what should run; a lift study validates what did run. Teams that get value from both sequence them instead of treating them as competing line items.
Take a $120,000 flight. Pre-testing six concepts on two to three percent of it costs less than a single one-question lift study, and happens while changing the ad is still cheap. Launch the two that scored, then put $10,000 of the live media behind a one-question ad-recall study to confirm the read in market. You have bought a decision and a validation for roughly what the validation alone would have cost.
That is the same logic behind copy testing and the broader practical guide to creative testing — and it is why scoring creative across dimensions gives you the diagnosis a lift number never will. To see a pre-flight read on your own creative, AdTest.AI scores it in about ninety seconds.
Frequently asked questions
How much does a brand lift study cost?
There is no separate fee — you pay in media. Google requires $10,000 in qualifying spend within 10 days for a single-question study in the US and the UK, rising to $60,000 for three questions. Meta publishes minimums per country. Agency or panel-based studies through firms like Nielsen or Kantar are priced separately and typically cost more.
How long does a brand lift study take?
Google’s studies usually complete within three to fourteen days once the spend threshold is met. The practical constraint is not the survey but the media: you need the flight running long enough and wide enough to build both groups.
What counts as a good lift?
It depends on the metric. Ad recall moves most readily and often shows lifts in the high single digits or better; purchase intent rarely moves more than a point or two, and frequently returns no statistically significant result at all. Compare against your own past studies rather than a published benchmark.
Can I run a brand lift study on a small budget?
Not below the platform minimums, and it is not worth trying to sneak under them. An underpowered study returns an inconclusive result that costs the same as a conclusive one. If the budget is tight, pre-test the creative instead and save the lift study for the flight that justifies it.
Is brand lift the same as conversion lift?
No. Both use holdout groups, but brand lift measures survey-reported memory and attitudes, while conversion lift measures actual conversion behaviour between exposed and held-out users. They answer different questions and are often run together.
