A brand lift study is a controlled market research experiment that measures how much an advertising campaign changes upper-funnel brand metrics such as awareness, ad recall, consideration, favorability and purchase intent. In marketing, it answers a question that clicks and conversions cannot: whether the campaign actually shifted what people think and remember about the brand.
The method compares people who were exposed to the ads with a comparable control group that was not exposed, and treats the difference in survey answers as the lift attributable to the campaign. Because it relies on randomization rather than self-reported memory alone, a well-designed study isolates the causal effect of advertising from background trends.
How a brand lift study isolates the effect of advertising
A brand lift study works by splitting the target audience before the campaign runs and surveying both groups with identical questions. The exposed group is eligible to see the ads, while the holdout or control group is deliberately withheld from them or shown an unrelated placeholder ad. Any systematic difference between the groups can then be attributed to the advertising itself.
The core output is absolute lift, the percentage-point difference between exposed and control respondents, and relative lift, the absolute lift divided by the control baseline. Analysts also report statistical significance and a confidence interval, because survey answers are noisy and small differences often arise by chance.
The following table describes the main design approaches and what each one controls for.
| Design | How the control is formed | Main strength | Main weakness |
|---|---|---|---|
| Randomized holdout | Users randomly withheld from ads | Strongest causal evidence | Requires platform support and lost reach |
| Ghost ads | Platform logs who would have seen the ad | No wasted media spend | Available only on some ad systems |
| Placebo ad (PSA) | Control sees an unrelated ad | Equalizes ad load between groups | Placebo media costs money |
| Matched exposed versus unexposed | Statistical matching after the fact | Works without pre-planning | Vulnerable to selection bias |
| Geo experiment | Whole regions held out | Captures offline and cross-channel effects | Needs large budgets and many regions |
Metrics measured in a brand lift study
Each metric in a brand lift study corresponds to a stage of the purchase funnel, from first noticing the brand to intending to buy. Most studies ask one question per metric, often only one to three questions in total, because longer surveys reduce response rates and add fatigue bias.
| Metric | Typical survey question | What a lift indicates |
|---|---|---|
| Ad recall | Do you recall seeing an ad for this brand recently? | The creative was noticed and remembered |
| Aided awareness | Which of these brands have you heard of? | The brand entered memory for new audiences |
| Unaided awareness | Which brands come to mind for this category? | The brand gained mental availability |
| Consideration | Which of these brands would you consider? | The brand joined the shortlist |
| Favorability | How do you feel about this brand? | Attitudes became more positive |
| Purchase intent | How likely are you to buy from this brand? | Movement toward conversion |
Lift tends to be largest for ad recall and smallest for purchase intent, because deeper attitudes move slowly. A campaign that produces 8-15 points of recall lift but only 1-2 points of intent lift is common and not necessarily a failure; the relevant benchmark depends on category, brand maturity and media weight.
Brand lift compared with related market research methods
A brand lift study differs from other research tools because it is experimental and campaign-specific. A general survey or brand tracker describes the market over time, while a brand lift study assigns causality to one campaign within a fixed window.
| Method | Primary question | Causal evidence | Typical timing |
|---|---|---|---|
| Brand lift study | Did this campaign change perceptions? | Yes, through a control group | During or right after a campaign |
| Brand tracking survey | How is brand health trending? | No, descriptive only | Continuous or quarterly |
| Conversion lift study | Did this campaign drive actions? | Yes, on behavioral outcomes | During a campaign |
| Focus group | Why do people react this way? | No, qualitative insight | Before or after a campaign |
| Marketing mix modeling | How do channels contribute over years? | Modeled, not experimental | Annual or semiannual |
In practice, these methods are complementary rather than competing. Focus groups and customer interviews explain the reasons behind a result, a brand lift study confirms whether a specific campaign worked, and marketing mix modeling places that campaign within the long-term budget picture.
Designing and running a reliable study
Reliable results depend on planning the study before launch, not after the campaign ends. The team should define one primary metric and a hypothesis in advance, because testing many metrics and reporting only the winner inflates false positives.
- Choose the primary metric that matches the campaign objective, such as ad recall for a launch or consideration for a challenger brand.
- Estimate the minimum detectable lift and the required sample, which often means several thousand completed responses per group.
- Fix the holdout share, commonly 5-20 percent of the audience, balancing statistical power against lost reach.
- Field the survey to both groups in the same time window with identical wording and answer order.
- Report absolute lift, relative lift, the confidence interval and the cost per lifted user.
The cost per lifted user, which is media spend divided by the estimated number of people whose answers changed, lets planners compare creatives and channels on a common scale. It converts a percentage into a budget efficiency figure that finance teams can evaluate.
Limitations and common mistakes
A brand lift study measures stated attitudes, not actual purchases, so a positive result does not guarantee sales. Survey answers are subject to response bias, and people who answer in-feed polls may differ from the wider audience.
| Problem | Cause | Remedy |
|---|---|---|
| Underpowered study | Too few responses for a small expected lift | Calculate sample size before launch |
| Contaminated control | Control users saw ads elsewhere | Use platform holdouts or geo splits |
| Leading questions | Wording hints at the advertised brand | Include competitor brands as answer options |
| Cherry-picked metrics | Many metrics tested, best one reported | Pre-register one primary metric |
| Short-term snapshot | Survey fielded only during the flight | Add a delayed wave to test decay |
Another frequent misconception is that platform-run studies are fully neutral. They are methodologically sound when randomized, yet each platform measures only its own inventory, so cross-channel comparisons require independent or geo-based designs.
In plain words
A brand lift study checks whether ads changed minds. Before the campaign starts, the audience is split into two similar groups. One group can see the ads and the other cannot. Afterward, both groups answer the same short question, such as whether they remember the brand.
If more people in the ad group remember or like the brand, that gap is the effect of the advertising. It works like a medical trial with a placebo group: the comparison shows what the ads achieved on their own, rather than what would have happened anyway.
A real-world example
A regional plant-based yogurt brand plans a four-week online video campaign with a budget of 120,000 dollars to reach adults aged 25-44. The marketing team runs a brand lift study with a 10 percent randomized holdout and chooses aided awareness as the primary metric, with consideration as a secondary one.
After the flight, about 6,000 respondents per group have answered. Aided awareness reaches 31 percent in the exposed group against 24 percent in the control group, an absolute lift of 7 points and a relative lift of about 29 percent. Consideration rises by only 1.5 points, which is not statistically significant.
The team concludes that the campaign built awareness successfully but did not yet persuade viewers to add the brand to their shortlist. For the next quarter, it keeps the reach-focused video and adds product-benefit creatives aimed at people who have already seen the brand, then plans a second study to test consideration.