Meta Brand Lift Study Questions: What They Measure and How to Use Them

The question types Meta actually offers

Meta doesn't let you write brand lift questions from scratch. You pick a construct from a fixed set, and Meta drops in your brand or product name. That matters because a lot of confusion about "what questions to use" comes from advertisers expecting a free-text survey builder, which this isn't.

• Brand awareness: "Have you heard of [Brand]?" — tests whether the ad put you on the map at all.
• Ad recall: "Do you recall seeing an ad from [Brand] in the last two days?" — the standard for reach and video campaigns.
• Favorability: "How do you feel about [Brand]?" on a positive-to-negative scale — used when the goal is perception repair or a rebrand.
• Consideration: "Which of these brands would you consider buying from?" shown against a competitor set — useful mid-funnel.
• Purchase intent: "How likely are you to buy from [Brand] in the next 30 days?" — the closest BLS gets to bottom-funnel, and the most sensitive to a mismatched audience.

Each of these is a randomized test-vs-control poll, not a click-through survey. The control group never saw the ad. That's the part worth sitting with: BLS is a holdout test wearing a survey costume.

How to pick the right question for the campaign you're actually running

The question has to match the funnel stage of the media, not the funnel stage you wish you were at. A broad-reach prospecting campaign with cold audiences should test awareness or ad recall. A retargeting or catalog campaign aimed at people who already know the brand should test consideration or purchase intent instead.

Get this backwards and the study tells you nothing useful. Ask a cold, top-of-funnel audience a purchase-intent question and most of them will say "unlikely" regardless of the ad, because they weren't close to buying in the first place. That reads as campaign failure. It's actually a measurement design failure — the question and the audience were never going to agree.

The mistake almost every advertiser makes here

We've seen this one made — and made it ourselves running mixed programs — pairing a single Brand Lift Study across several campaigns with different creative, different offers, and different audiences, then asking one awareness question across all of it. Meta requires consistent creative theme and consistent targeting across the campaigns in a study for exactly this reason, and it's the rule advertisers skip most.

Do it anyway and the lift number that comes back is an average of several different truths. A strong-performing hero creative and a weak filler creative get blended into one flat result, and the flat result gets blamed on "brand lift doesn't work" rather than on the setup.

The second common error is stacking too many question constructs into one study to save budget. Response rates on in-feed polls are already low. Split that thin sample across three or four question types and none of them reach statistical confidence. One clean question beats three noisy ones.

What to do when the lift comes back flat or negative

Before concluding the creative failed, check the setup in this order:

• Sample size: did the campaign actually clear the minimum impressions and poll responses Meta needs for a statistically significant read? Thresholds vary by market and Meta changes them without much notice — check current eligibility inside Ads Manager rather than trusting a number from last year's guide.
• Test/control contamination: if the same person could plausibly see the ad through another campaign or another channel, the control group isn't clean anymore, and the lift gets diluted toward zero.
• Timing: lift studies need time for survey responses to accumulate on both sides. A read pulled too early looks flat because it's incomplete, not because it's negative.
• Question-to-objective match: re-check the previous section. A wrong question on a right campaign produces a wrong-looking result.

Only after ruling those out is it fair to say the creative or targeting genuinely didn't move perception. That's a real, useful finding — it just has to survive the checklist first.

Why this matters beyond Meta

The reason BLS exists at all is the same reason last-click attribution keeps getting challenged everywhere else: exposure to an ad and response to an ad get conflated constantly, and only a randomized control group separates the two. Meta built that separation into BLS by default. Most advertisers don't get the equivalent on other channels unless they build it themselves.

On Amazon, that gap is why last-click sponsored ads and DSP reporting double-count the same shopper, and why holdout tests and clean-room reconciliation in Amazon Marketing Cloud exist — same logic as a BLS test-and-control split, applied to a different funnel. Across 30 advertisers we managed in July 2026, that kind of measurement produced a blended $5.49 cost per acquisition across 57,137 attributed purchases, 20.1% of them new-to-brand — numbers we trust specifically because they came from holdout-based measurement, not from a last-click report crediting itself twice.