Meta brand lift study: setup, minimum budgets and how to read results

Updated October 2026.

A brand lift study answers the question that reach and CPM reports can’t: did your ads make more people remember your brand? In Vietnam, running a self-serve Brand Lift test in the Experiments tool of Meta Ads Manager requires a campaign budget of at least $29,000 (the minimum Meta has applied since September 15, 2026). This guide covers eligibility, the setup steps, how to choose survey questions, how to read the results with a worked example, and a checklist to run through before you launch.

It does not debate whether brand investment is worth it. Every threshold, button name and formula below comes from the Meta Business Help Center and Google Ads Help, checked on October 27, 2026.

What a Meta brand lift study is and how Meta measures it

Meta selects a representative sample of people eligible to see your ads and splits it at random into two groups. The test group has the opportunity to see your ads. The holdout group is held back for the whole test. Both groups get the same survey question, and the difference in response rates is the lift your ads caused (Meta, About Brand Lift tests).

Think of it like this: you hand out flyers to half of a neighborhood and skip the other half. A week later you ask everyone, “Have you heard of Cafe X?” However many more “yes” answers come from the flyer half is what the flyers achieved. People who already knew the cafe live in both halves, so they cancel out.

The survey appears in Facebook Feed and Instagram Feed. Topics you can ask about include ad recall, brand awareness, favorability, familiarity, recommendation, consideration and purchase intent (Meta, About brand lift poll questions).

Brand Lift vs. the estimated ad recall lift metric

Brand Awareness campaigns (the Awareness objective with the ad recall lift performance goal) already report a metric called Estimated ad recall lift (people). Meta produces this number with machine learning, calibrates it with surveys run on a random sample of campaigns, and labels it “estimated” and “in development.” Meta recommends comparing it only across campaigns running at the same time, and says a Brand Lift test is still the most accurate way to measure ad recall (Meta, About the ad recall lift metric).

When to run a brand lift study, and when not to

Meta recommends Brand Lift when you run one campaign for one brand or product and keep the creative and budget the same throughout. It also fits when you can’t measure conversions directly (Meta, About brand survey tests). This is brand awareness measurement with a control group, so it only pays off when the campaign is large enough and clean enough.

• Run it for: product launches, brand identity campaigns when you change a logo or packaging, and big seasonal awareness pushes; products sold through supermarkets or distributors, where sales never reach your Pixel.
• Hold off when: the same brand has other campaigns live on Meta. People in the holdout group may see ads from those campaigns and contaminate the results.
• Wrong tool: you want to know how many extra orders your ads drove. That is a job for Conversion Lift.

Minimum budget in Vietnam

Meta sets a minimum budget per country so that the test collects enough responses for statistical significance. If the campaign doesn’t meet it, you can’t create a Brand Lift test in Experiments. You can run the test on one campaign that meets the minimum, or across the whole ad account if total spend meets it (Meta, Brand Lift minimum requirements).

If you target several countries, the system applies the highest minimum: Vietnam plus Malaysia means $44,000. Ads Manager converts the figure into your account currency at Meta’s exchange rate, so the number you see may differ slightly from the table. Meta notes these minimums can change, so check the page again before you plan.

The reliable check is to open Experiments, click Create test and see whether Brand Lift is available for your account.

Duration, response count and the holdout group

Meta recommends running the test for at least 4 weeks to collect more than 250 survey responses, and it can’t overlap with another test (Meta, Create a brand survey test). In a single-cell test (one group of campaigns), the holdout group is 10% the size of the test group, and Meta scales it up to the size of the test group when comparing (Meta, View and understand brand survey test results). The holdout group can push CPM up, because some people in your target audience are kept from seeing the ads.

Self-serve in Experiments or through a Meta rep

Neither route charges for the test itself; you pay only for the ads. For Meta-managed tests, Meta says eligibility depends on multiple factors and availability isn’t guaranteed (Meta, Create a Meta-managed Brand Lift test).

How to set up a Brand Lift test in Ads Manager

Create the campaign first, with a budget that meets the minimum. Ad set schedules must fall completely inside or completely outside the test dates: if the test runs June 8 to July 8, don’t create an ad set that runs June 1 to June 15.

In Ads Manager, open the navigation on the left, select All tools, then select Experiments under Analyze and report (Meta, How to set up a conversion lift test). Then follow Meta’s sequence (Meta, Create a brand survey test):

1. Click Create test at the top, find Brand lift and click Get started.
2. Under Learning objectives, choose the purpose of the test.
3. Choose Account (all ads in the ad account) or Campaign (one campaign). If you pick Campaign, Meta attaches that campaign’s Page to the survey automatically.
4. Pick a region and a vertical (industry) for comparison. The report will place your lift next to the median for that region and industry.
5. Set the test start and end dates (separate from the campaign dates) and a test name.
6. Choose a survey language that matches your audience. Only people who use Facebook in that language can see the survey; for an audience in Vietnam, choose Vietnamese (vi_vn).
7. Create up to 3 questions.
8. Choose whether to measure Conversion Lift in the same test. It doesn’t affect Brand Lift, but the account must already be measuring conversions.
9. Click Review test details, check everything, then click Create test.

Some Pages can’t run brand surveys in Experiments, depending on the kind of brand. Alcohol brands, for example, can’t in certain locations. Beer and spirits advertisers should check this first.

How to write brand lift study questions

The questions decide what a Brand Lift test can tell you, so this is the step of the brand survey worth the most time. Every test must include the standard ad recall question. Each person sees one question at a time, but may see the others over the course of the test. Meta’s templates in Experiments (Meta, About brand survey poll questions):

Meta-managed surveys add a top-of-mind question (“Which of the following comes to mind first when you think of [brand/product]?”) and a purchase intent question (“Will you buy/consider [brand/product] the next time you shop for [category]?”).

The formula I use to pick questions: [required ad recall] + [1 question matching the campaign goal] + [1 question one step further down the funnel].

• New brand at launch: ad recall + brand awareness + familiarity.
• Established brand changing its image: ad recall + favorability + recommendation.
• Working with a Meta rep before a shopping season: ad recall + awareness + purchase intent.

A well-known brand has little room to grow on awareness. If 90 out of 100 people already know you, your ads have only 10 people left to win over. In that case, favorability or recommendation tells you more.

What to avoid when writing questions

Experiments only lets you edit a few words through dropdowns, so wording mistakes mostly happen in Meta-managed surveys or when you fill in product and category names.

• Leading questions: “Do you love X’s delicious new peach flavor?” The question suggests its own answer. Keep it neutral: “How would you describe your overall opinion of X?”
• Two questions in one: “Have you heard of X and do you plan to buy it?” Someone who knows the brand but has no plan to buy can’t answer.
• Categories that are too narrow: “next time you shop for peach drinkable yogurt.” Use the category shoppers actually think in: “drinkable yogurt.”
• Questions touching sensitive topics such as health, finances or religion. Meta-managed questions all go through Meta’s approval, and these are likely to be rejected.

How to read brand lift study results

Go to Experiments, click Learn, find your test and click View report. The report has three sections: Results (lift summary), Poll questions (each question, broken down by age and gender) and Test details (schedule, campaigns, creative and spend distribution) (Meta, View and understand brand survey test results).

1. Brand lift percent, or absolute lift = the desired response rate in the test group minus the rate in the holdout group, in percentage points. Meta’s example: 124 out of 512 (24.22%) minus 63 out of 498 (12.65%) gives 11.6 points (Meta, Brand lift percent). A negative number is usually statistical noise.

2. Relative lift = absolute lift / holdout group rate. In the example above: 11.6 / 12.65 = 91.7%. Meta’s report shows lift in percentage points; I calculate relative lift myself to compare questions with different baseline rates.

3. Brand lift = accounts reached x brand lift percent. Meta’s example: 1,830,000 accounts x 11.6 points, roughly 212,000 accounts that remembered the brand because of the ads (Meta, Brand lift). This number follows the campaign’s reach and frequency.

4. Cost per brand lift = spend / brand lift, often called cost per lifted person. Meta’s example: $200,000 / 212,000 = $0.94 per additional account that remembered the ad (Meta, Cost per brand lift). When you compare two strategies, the one with the lower cost per brand lift is usually the more effective one for awareness.

5. Confidence: for lift tests, 90% or higher is a statistically reliable result. Each question has its own confidence (Meta, About confidence in your tests).

Step-by-step example (hypothetical numbers)

Say I’m a drinkable yogurt brand launching a peach flavor, targeting people aged 18 to 34 in Vietnam. Every number below is hypothetical, used only to show the math.

Step 1. Budget. A 4-week campaign, $32,000, targeting Vietnam only, above the $29,000 minimum. I pause the brand’s other Meta campaigns for those 4 weeks.

Step 2. Questions. It’s a new product, so I pick ad recall (1), brand awareness (2) and familiarity (3).

Step 3. Results after 4 weeks. The campaign reached 3,000,000 accounts:

Step 4. Calculate for question (1). Brand lift = 3,000,000 x 6.0% = 180,000 accounts. Cost per brand lift = 32,000 / 180,000 = $0.18.

Step 5. Conclusion. Questions (1) and (3) are at 90% or higher: the ads produced lift. Question (2) is only at 68%, so no conclusion is possible. 58% of the holdout group already knew the parent brand, so there was little room to grow. Next time I’d swap the awareness question for favorability, then compare each question’s lift with the regional and industry medians in the report.

Brand lift results calculator

Copy this into Google Sheets, one column per question (B for question 1, C for question 2 and so on). Rows marked “Input” come from the report.

Your own numbers may differ slightly from the report, because Meta scales the holdout group and uses causal inference. When they differ, go with Meta’s figures.

Common brand lift mistakes

1. Budgeting right at the minimum, then cutting mid-test. Underspending makes it hard to pass 250 responses, or leaves every question below 90% confidence.
2. Editing the campaign while the test runs. Meta advises changes only to fix problems, not to optimize. Changing creative, audience or budget makes the results less meaningful.
3. Overlapping. A test can’t overlap with another test. Other campaigns for the same brand will show ads to the holdout group and drag lift down.
4. Ad sets that straddle the test dates. Spend and exposure no longer line up with the survey period.
5. Drawing conclusions at 250 responses. Results start appearing at that point, but Meta recommends waiting until the test ends.
6. Comparing estimated ad recall lift across periods. Meta keeps changing how it is calculated, so comparing March with September tells you nothing.

Brand Lift vs. Conversion Lift and incrementality testing

Brand Lift is a form of incrementality testing: it compares people who could see your ads with people who couldn’t, to isolate what the ads actually caused. The difference is what gets measured.

For Conversion Lift, Meta recommends at least $5,000 and at least 28 days (Meta, About conversion lift). A/B tests have no holdout group, so they can’t measure incremental impact (Meta, About lift and holdouts).

Brand Lift gives you one clean number for one campaign on one platform. It doesn’t replace Marketing Mix Modeling (MMM) for estimating every channel’s effect on sales, or multi-touch attribution at the touchpoint level. Put brand measurement results next to growth marketing metrics such as MER and CAC for the same period, to see whether higher awareness came with more new customers. For apps, see iOS app ad measurement.

Google Ads and YouTube Brand Lift

Google offers free Brand Lift for Video and Demand Gen campaigns. Surveys show on YouTube before a video starts. Reports include lifted users, cost per lifted user, absolute brand lift, relative brand lift and headroom brand lift (Google Ads Help, About Brand Lift). Unlike Meta, Google requires you to work with a Google account representative, the campaign must meet a minimum spend within 10 days (Google Ads Help, About Brand Lift), and surveys take up to 48 hours to be approved (Google Ads Help, Set up Brand Lift). Don’t add Meta’s lift to Google’s: different questions, different respondents, different math.

Brand lift study checklist before launch

Answer each item yes or no. Any “no” means don’t click Create test yet.

1. The budget meets the minimum for the most demanding country in the audience, checked against the latest Brand Lift minimum requirements page.
2. Brand Lift is available in Experiments for the account.
3. Only one campaign for this brand or product runs on Meta during the test.
4. No other test overlaps.
5. Ad set schedules fall completely inside or outside the test dates.
6. The test runs for at least 4 weeks.
7. Creative and budget are locked.
8. The survey language matches the audience (vi_vn for Vietnam).
9. Questions follow the formula ad recall + 1 goal question + 1 next-step question, with no leading wording.
10. Region and vertical are set correctly for the benchmarks.
11. Decision thresholds are written down in advance: 90% confidence or higher, and the cost per brand lift at which you scale.
12. The team knows CPM may rise because of the holdout group and won’t re-optimize because of it.

When you hand many creatives to Meta Andromeda to distribute, the Test details section shows where spend went by creative, next to the lift for the whole campaign. If you need to plan an awareness campaign that qualifies for Brand Lift and read the results with your team, contact Adtek.

Sources

• Meta Business Help Center (accessed October 27, 2026). About Brand Lift tests; About brand survey tests.
• Meta Business Help Center (minimums as of September 15, 2026, accessed October 27, 2026). Brand Lift minimum requirements.
• Meta Business Help Center (accessed October 27, 2026). Create a brand survey test; About brand survey poll questions in Experiments; About brand lift poll questions; Create a Meta-managed Brand Lift test.
• Meta Business Help Center (accessed October 27, 2026). View and understand brand survey test results; Brand lift percent; Brand lift; Cost per brand lift; About confidence in your tests and experiments.
• Meta Business Help Center (accessed October 27, 2026). About the ad recall lift metric; About Experiments; About lift and holdouts; About conversion lift; How to set up a conversion lift test.
• Google Ads Help (accessed October 27, 2026). About Brand Lift; Set up Brand Lift.
• Google Ads Keyword Planner, Vietnam market, Vietnamese language, checked October 27, 2026.