How to Measure Brand Lift: Absolute vs Relative Lift and Study Design

Published:
Last Updated:
Category: Marketing Glossary, Advertising Operations
Authors: Shusaku Yosa
Brand lift evaluates changes in awareness or attitudes using a comparison condition. It asks different questions from impressions or clicks: do people recognize the brand, remember the advertising, or consider choosing it? A difference between people who saw an ad and those who did not is not automatically a causal advertising effect.
Define the intended change, relevant audience, and comparison before collecting results. Report percentage-point differences separately from relative increases, with the base rates and uncertainty. The calculation is useful only when its interpretation matches the study design.
Choose one primary question
Brand awareness, ad recall, consideration, and purchase intention are distinct outcomes. Decide which one the campaign should influence and specify the question and positive response before launch. Do not combine consideration with purchase intention simply because both concern a future choice.
Audience, product, and timing affect answers. An immediate purchase count may not capture the purpose of an awareness campaign, while awareness growth alone does not establish revenue growth. Separate a continuing brand diagnosis from the evaluation of a specific intervention.
Understand how the comparison was created
People reached by advertising may already be more interested or use different media. Lewis, Rao, and Reiley’s advertising measurement research discusses experimentation and causal measurement. Matching observed characteristics after delivery can still leave unobserved differences.
Comparison | Check | Do not assume |
|---|---|---|
Randomly assigned treatment and control | Assignment, delivery, other exposure, and missing responses | Reselecting only actual viewers preserves randomization. |
Observed exposed and unexposed people | Prior interest, media behavior, and audience composition | The observed difference is incremental ad impact. |
Before and after | Respondent composition, seasonality, competitors, and price | Only the campaign changed the outcome. |
In an experiment, assignment and actual exposure remain different. Excluding people who were not reached or did not answer can change the quantity being estimated. Specify the analysis before looking at results. When only an observational comparison is feasible, label it accordingly and document remaining bias.
Distinguish percentage points from relative lift
Consider this fictional calculation, not a real study. In two comparable groups, 40 of 200 control respondents and 50 of 200 treatment respondents answer positively. The rates are 20% and 25%. Google Ads’ metric documentation also explains absolute and relative lift terminology.
Quantity | Calculation | Result |
|---|---|---|
Control positive response rate | 40 ÷ 200 | 20% |
Treatment positive response rate | 50 ÷ 200 | 25% |
Absolute lift | 25% − 20% | 5 percentage points |
Relative lift | (25% − 20%) ÷ 20% | 25% |
Writing only “5% improvement” is ambiguous. State “20% to 25%, a five-percentage-point difference and 25% relative lift.” When the control rate is zero, this relative-lift formula is undefined. A low baseline can also make a relative figure appear dramatic, so retain the underlying rates.
Read precision alongside the point estimate
The fictional five-point difference is a point estimate. Two hundred responses per group do not automatically make it conclusive. Review the uncertainty interval, estimation method, and decision criterion for the design. An interval including zero does not support reading the result as an exclusively positive change.
Insufficient data and no detected lift are different states. Non-detection does not prove an effect is exactly zero; the study may not distinguish it clearly at the available precision. Detailed audience breakdowns reduce sample sizes, making it risky to celebrate the largest subgroup result.
Specify primary outcomes, dates, exclusions, and planned subgroup analyses in advance. Adding many cuts after seeing results increases the opportunity to find a convenient difference. Label exploratory findings as hypotheses for later testing. A platform report and an independent survey may also use different definitions or estimators.
Use the outcome pattern to choose a hypothesis
If awareness changes but consideration does not, investigate product relevance and the explanation people receive after noticing the brand. If ad recall is high but brand recognition is weak, the creative may be memorable without a strong brand connection. Treat these as hypotheses and check comprehension or open-ended responses.
Higher responses among frequently exposed people do not establish that increasing frequency will reproduce the same gain. Interested people may have received more exposure. Separate causal evaluation from exploratory diagnosis, and avoid projecting a survey difference directly into revenue or budget returns.
Keep decision conditions in the report
- Market, audience, dates, intervention, and control conditions
- Question wording, positive response definition, group sample sizes, and rates
- Absolute and relative units, uncertainty intervals, and decision criteria
- Missing data, other campaigns, potential bias, and the next hypothesis
Use brand equity for the broader diagnosis, brand loyalty for attitudes and repeat choice, and KPI progress decisions for the operating response. A useful lift report explains both the difference and the conditions under which it was obtained.




