MEDUSA

Measurement & Decisioning · 22 August 2026 · 7 min read

How to measure brand lift: methods, sample size and common mistakes

Brand lift measures whether advertising changed what an audience knows, remembers or feels about a brand. The strongest studies compare a randomly assigned exposed group with a comparable control group, then measure the difference in awareness, recall, consideration or intent with enough responses to distinguish a real effect from sampling noise.

Key takeaways

  • Choose one primary brand metric before launch and tie it to the campaign objective; do not shop across survey questions for the most flattering result.
  • Exposed-versus-control designs are stronger than simple pre-versus-post comparisons because they estimate what would have happened without the campaign.
  • Sample size depends on the size of lift you need to detect: smaller effects require substantially more responses.
  • Report absolute lift, relative lift, the uncertainty range and the study design together. A percentage without its base and confidence is not a planning result.

What brand lift means

Brand lift is the difference in a brand metric between people who had the opportunity to see a campaign and a comparable group who did not. Typical metrics include ad recall, awareness, message association, consideration, favourability and purchase intent. Unlike clicks or attributed conversions, these measures are designed to read changes in perception that happen before an immediate action.

The metric should follow the campaign objective. A launch intended to introduce an unknown product might prioritise aided awareness. A campaign communicating a new proposition might use message association. A mature brand trying to enter the consideration set may measure consideration or preference. Selecting all of them without a primary outcome increases the chance of finding a positive result by accident.

The strongest design: exposed vs control

A randomised brand lift study creates two statistically comparable groups. The test group is eligible to receive the campaign; the control group is held back. Both groups receive the same survey. Because randomisation balances many other influences, the difference in answers can be attributed more credibly to the advertising than a simple before-and-after change can.

MethodWhen it is usefulMain limitation
Platform-native randomised liftLarge campaigns on a platform with built-in holdout and survey deliveryReads that platform's eligible audience and methodology
Independent panel studyCross-channel campaigns or questions requiring custom audience and survey designExposure matching and panel quality require careful validation
Geo experiment plus brand trackingCampaigns that can vary media pressure across comparable marketsNeeds enough similar geographies and control of spillover
Pre/post surveyDirectional tracking when a holdout is impossibleCannot cleanly separate media from news, seasonality or other changes

How to choose the right brand lift metric

Campaign jobPrimary metricExample survey question
Create memoryAd recallWhich of these brands have you seen advertised recently?
Build salienceAided or unaided awarenessWhich brands in this category come to mind?
Land a propositionMessage associationWhich brand do you associate with this claim?
Enter the shortlistConsiderationWhich brands would you consider?
Move preferenceFavourability or intentHow likely are you to choose this brand?

Write the hypothesis before the study is configured: 'Among category buyers aged 25-44, the campaign will increase aided awareness by at least four percentage points.' That sentence fixes the audience, outcome and decision threshold. It also makes the later read honest: if the lift is smaller than the level that would change the plan, statistical significance alone does not make it commercially important.

Sample size and statistical power

Brand lift studies are frequently underpowered. The detectable effect depends on the baseline response rate, expected lift, confidence level, statistical power and the split between exposed and control respondents. The smaller the effect, the larger the sample required. A study designed to detect a four-point lift may be unable to distinguish a real one-point lift from noise.

Run the sample-size calculation before launch, using the minimum commercially meaningful effect rather than an optimistic expected result. Platform thresholds can change and differ by market, format and objective, so confirm current eligibility and response guidance. Google Ads' Brand Lift guidance illustrates how required responses rise sharply as the detectable absolute lift becomes smaller.

How to calculate and report brand lift

Suppose 40% of control respondents recognise the brand and 46% of exposed respondents do. Absolute lift is 46% minus 40% = 6 percentage points. Relative lift is 6 divided by 40 = 15%. Headroom lift compares the gain with the maximum improvement available: 6 divided by the 60-point gap between the control result and 100%, or 10% of headroom.

Illustrative awareness result
MeasureCalculationResult
Control awarenessSurvey result40%
Exposed awarenessSurvey result46%
Absolute lift46% - 40%+6 percentage points
Relative lift6 / 40+15%
Headroom lift6 / (100 - 40)10% of available headroom

Always report the confidence interval with the point estimate, plus field dates, audience, markets, sample size, allocation method, survey question and any exclusions. 'Brand awareness increased 15%' is ambiguous; 'aided awareness rose by six percentage points, from 40% to 46%, with this uncertainty range' is interpretable.

How planners should use the result

A positive total-campaign lift does not automatically identify which channel or creative deserves more budget. Breakouts can generate hypotheses, but they need adequate sample in every cell. Use the study first to answer the question it was powered for, then combine it with reach, frequency, share of search, incrementality and delivery evidence before changing the mix.

Result patternLikely interpretationPlanning response
Recall up, message association flatCreative was noticed but the proposition did not landFix message clarity before adding reach
Awareness up, consideration flatCampaign created salience but not preferenceReview proposition, proof and mid-funnel role
Strong lift, low incremental reachA responsive group moved but scale was limitedTest broader reach without losing relevance
No detected lift, low powerStudy cannot distinguish effect from noiseDo not call the campaign ineffective; redesign the measurement
No detected lift, adequate powerThe planned effect was not achievedDiagnose audience, reach, frequency and creative before repeating spend

Common brand lift mistakes

  1. Choosing the question after seeing the result. Pre-register the primary metric and decision threshold.
  2. Using a weak control. A mismatched or contaminated control group makes the difference hard to attribute to media.
  3. Cutting the sample too many ways. Market, age, platform and creative breakouts quickly turn a powered total into dozens of noisy cells.
  4. Confusing percentage points with percent. A move from 40% to 46% is six points and 15% relative lift; report both labels correctly.
  5. Ignoring frequency and reach. A campaign cannot move a broad audience if too few people had a credible opportunity to receive the message.
  6. Treating non-significance as proof of no effect. It may mean no effect, or simply not enough information to detect one.
  7. Optimising only to the study. Brand lift is one planning signal; it does not replace sales, incrementality, cost and long-term demand evidence.

Frequently asked questions

What is brand lift?

Brand lift is the measured change in a brand metric - such as awareness, recall, consideration or intent - associated with exposure to an advertising campaign, usually estimated by comparing exposed and control groups.

How is brand lift calculated?

Absolute lift is the exposed group's positive response rate minus the control group's rate. Relative lift divides that absolute difference by the control rate. Report the underlying rates, percentage-point difference and uncertainty so the result cannot be misread.

How many responses does a brand lift study need?

It depends on the baseline response, minimum effect to detect, confidence, power and group allocation. Smaller expected effects need much larger samples. Calculate the requirement before launch and confirm the current eligibility rules of the platform or research provider.

What is the difference between brand lift and conversion lift?

Brand lift measures changes in perception, usually through surveys. Conversion lift measures incremental actions such as purchases or leads, usually through behavioural outcomes in a holdout experiment. They answer different stages of the measurement problem.

Can brand lift be measured without a control group?

A pre/post survey can show directional change, but it cannot isolate the campaign from seasonality, news, competitor activity or other events as well as a valid control. Label it as tracking rather than causal lift.

Keep reading