How to Design a Brand Lift Survey: Groups, Questions, and Reporting
Published:
Last Updated:
Category: Marketing Glossary, Advertising Operations
Authors: Shusaku Yosa
A brand lift survey examines differences in awareness, recall, or consideration associated with advertising. Before writing questions, decide what you need to learn, which groups you will compare, and what the design can establish about the advertisement's influence. A useful brief specifies the comparison, questionnaire, sampling approach, and reporting rules.
1. Choose a primary question about the brand
“Did the brand become stronger?” is too broad. Decide whether the priority is recognition of the brand name, recall of an advertisement, or consideration for a future purchase. Select a primary measure and keep related measures secondary. Purchase intention is an answer about intention, not a count of purchases.
Google's overview of lift studies distinguishes brand perception, search behavior, and conversions as different outcomes. Do not convert search activity into an awareness rate or treat positive survey answers as sales.
2. Specify the groups and treatment of exposure
People who saw an advertisement may already differ from people who did not. A randomized holdout experiment is therefore different from recruiting exposed and unexposed respondents afterward. In an observational comparison, matching measured characteristics does not eliminate every unmeasured difference.
Describe who is eligible for assignment, how advertising is withheld where applicable, how actual exposure is recorded, and how other campaigns could affect the comparison. Assignment and actual exposure are different concepts. Agree on the analysis approach before viewing results, including whether analysis follows original assignment.
Planning item | Fictional Brand A example | Question to resolve |
|---|---|---|
Primary objective | Brand awareness among eligible people | Keep awareness distinct from recall |
Population | People meeting stated regional and audience conditions | Are existing customers included? |
Comparison | Groups defined before the campaign | How are assignment and exposure handled? |
Timing | A planned two-week period and survey date | What other major activity occurs? |
Missing data | Unanswered items recorded separately | What are the exclusion and duplicate rules? |
3. Ask clear, non-leading questions
For fictional Brand A, a recognition question could ask which listed brands the respondent knew before the survey. Describing A as “popular and convenient” within the question would introduce an evaluation. Ask consideration separately and consider an appropriate answer for someone with no plans to use the category.
Specify response options, their order, whether multiple selections are allowed, and how “do not know” is treated. Pretest whether people understand the question. These are illustrative questions, not a promise that a platform allows unrestricted questionnaire editing. AAPOR's survey best practices provide a useful reference for question clarity and transparent reporting.
4. Plan recruitment and the precision needed
Reaching an eligible audience and obtaining representative responses are separate challenges. Examine who is likely to answer and whether recruitment or timing differs between groups. Record the population, recruitment method, achieved sample, and any weighting so another reader can assess the comparison.
The required sample depends on the difference you want to detect, the baseline rate, assignment, expected response, and other design choices. There is no universal rule that 500 respondents is sufficient. Platform availability also matters: Google's Brand Lift documentation states that the feature is not available for every advertising account. Confirm a suitable design with the provider before committing.
5. Report counts and percentage-point differences
This fictional example compares positive awareness responses. The sample sizes illustrate arithmetic; they are not recommended sample requirements or a substitute for assessing uncertainty.
Group | Respondents | Positive answers | Positive response rate | Difference |
|---|---|---|---|---|
Comparison group | 500 | 150 | 30% | Baseline |
Exposed group | 500 | 200 | 40% | 10 percentage points |
The absolute difference is 40% minus 30%, or ten percentage points. Relative to the comparison rate, the difference is (40 − 30) divided by 30, approximately 33.3%. A relative difference cannot be calculated this way when the comparison rate is zero. Report uncertainty using an approach appropriate to the design, alongside group definitions and counts. These answers do not establish an increase in purchasing.
6. Turn the finding into a defensible decision
A small estimated difference, an inadequate response count, and a disrupted comparison are different problems. Assess intervals or tests appropriate to the design. A result that is not statistically significant does not prove the effect is exactly zero; consider what range of differences the study could reasonably examine.
Report the primary measure first. Label additional segments or measures selected after examining results as exploratory. Then specify a decision: revise the creative, investigate the audience, or improve the measurement design. If you combine the survey with behavioral reports, preserve the population and unit of each measure rather than merging them into one success figure.
Related reading
