Skip to content

guide

Brand lift studies: meaning and how to run one

What a brand lift study is, how Google and TikTok run theirs, how to calculate lift, how many responses you need, and how to measure lift on radio and TV.

Takwimu editorial team, Editorial team8 min read

A brand lift study measures how much an advertising campaign changed what people think, by comparing survey answers from people who saw the adverts with similar people who did not. Digital platforms do this by holding back a random control group. Radio and TV cannot, so they need a design that owns its weakness.

Key points:

  • Brand lift compares an exposed group with a control group on the same questions.
  • Absolute lift is the difference in points. Relative lift flatters small brands.
  • Google and TikTok run this inside their own systems.
  • Broadcast lift usually compares places, or times, or recalled listeners.
  • Small samples can hide real lift and invent false lift, so plan the sample first.

What does brand lift mean?

Brand lift is the change in a brand measure caused by advertising. The measures are usually ad recall, awareness, consideration, favourability and purchase intent.

The word "caused" does the work. If awareness rises from 20 to 30 per cent during a campaign, some of that may be the season, a rival's mistake or a news story. A lift study tries to isolate the part the advert did.

How does a brand lift study work?

Two groups answer the same question. Google's Help Center describes showing surveys to people who saw the ads and to people who were eligible but were not shown them. The difference between the two gives the lift.

TikTok's help page describes the same idea: two randomised groups from part of the campaign's target audience, one exposed and one held back, compared through polling and statistics. Display & Video 360 calls the held-back people the control group.

Randomisation is the strength here. Because a coin toss decides who sees the advert, the two groups should differ only in that one thing.

How is brand lift calculated?

Use the formulas on Google's metrics page. Absolute lift subtracts the baseline group's positive rate from the exposed group's, and relative lift divides that gap by the baseline. Headroom lift divides absolute lift by one minus the baseline rate.

Here is an invented example with 30 per cent positive among the exposed and 22 per cent among the control:

Metric Sum Result
Absolute lift 30 minus 22 8 points
Relative lift 8 divided by 22 36.4 per cent
Headroom lift 8 divided by 78 10.3 per cent

Report absolute lift first. Agencies like relative lift because a brand starting at 2 per cent can post a large percentage from a small gain. Lifted users and cost per lifted user then scale the result to the campaign.

What are good brand lift study questions?

Ask one thing each, in plain words, in the language of the audience:

  • Ad recall: Which of these brands have you seen or heard advertised in the last two weeks?
  • Awareness: Which of these brands of seed have you heard of?
  • Consideration: Which of these would you consider buying this season?
  • Favourability: How do you feel about this brand, from very unfavourable to very favourable?
  • Purchase intent: How likely are you to buy this brand in the next three months?

Use a list with rivals on it. A question that names only your brand invites people to agree.

Can you run a brand lift study on Meta, YouTube, TikTok or LinkedIn?

Where the platform offers it, usually through a representative. Google says Brand Lift is not available on all accounts and that you need an account representative. TikTok says its studies can only be set up by your account manager.

Ask before you plan. Check each platform's own help pages for the terms that apply in Kenya, since a small campaign may not qualify.

Treat platform results as the platform's own marking. They measure the campaign inside that platform's walls, and they cannot compare it with what your radio spend did.

How many responses does a brand lift study need?

More than most people think. Display & Video 360's guide says about 1,200 to 2,800 responses can detect lift above 4 per cent, 5,000 to 11,000 can detect 2 per cent, and 20,000 to 45,000 can detect 1 per cent.

Survey-based broadcast studies face the same arithmetic. With an invented 400 people in each group, and the 30 and 22 per cent above, the 95 per cent margin on the 8-point gap is about 6.1 points. The lift is likely real, somewhere from 2 to 14 points.

Drop to 150 in each group and the margin grows to about 9.9 points. The same 8-point gap is now within chance. Weighting widens these margins further, and our page on sample size shows how.

How do you measure brand lift on radio and TV?

You cannot hold back a random group from a transmitter, so you choose among three weaker designs.

Compare places. Run the campaign in two counties and hold it back in two similar ones, then survey all four before and after. This is the strongest broadcast design, and it costs money in foregone reach.

Compare times. Survey the same county before and after. It is simple, and anything else that happened in those weeks gets credited to the advert.

Compare recall. Survey once, then split people by whether they recall listening to the station. It is cheap, and it carries a serious flaw.

What is wrong with comparing listeners with non-listeners?

They differ before the advert airs. People who like a station may already like the brands it carries, and users of a brand notice its adverts more. Recall of the advert is partly a result of liking the brand.

So exposure measured by memory runs the wrong way round. A group that recalls the advert may score higher because they already favoured the brand. Michelon and colleagues note that survey studies are likely affected by memory bias.

Use station listening over the advert itself as the split, measure before and after, and match the groups on age, gender and county. Say plainly what the design cannot rule out.

What does a broadcast brand lift study look like?

Here is an invented plan. A seed brand runs radio in two counties and holds back two similar ones.

Step What happens
Week 0 Survey 400 farmers in each county
Weeks 1 to 8 Radio runs in test counties only
Week 9 Survey 400 again in each county
Reading Change in test counties minus change in control counties

If awareness rises 12 points in the test counties and 3 in the control counties, the estimate is 9 points. That difference-in-differences reading removes the season and shared news.

Counties are the right unit here, because county census totals support the weighting. Our page on weighting an audience survey explains why.

When is a brand lift study worth doing?

When the decision is large enough. A study makes sense before renewing a big sponsorship, moving budget between media, or defending a brand budget to a board.

A small two-week burst rarely justifies it. The likely effect is smaller than the sample can see, and the study becomes an expensive way of learning nothing.

Commercial firms sell this work too: GeoPoll lists post-campaign evaluation among its tools. Ask any supplier for the control design and the margin of error before you buy.

What do advertisers get wrong about brand lift?

They run the survey after the campaign. With no baseline, there is no lift to report, only a level.

They read ad recall as success. Recall shows the advert was noticed. Consideration and intent are closer to the sale, and they move less.

They also stop at one number. A flat result overall can hide a strong result in one county, which a cross-tab shows at once. Our page on sponsorship evaluation applies the same thinking to a programme.

How does brand lift fit with reach?

Reach sets the ceiling. If 40 per cent of your target was reached and exposed people lifted by 8 points, the campaign moved about 3 points across the whole target. Lift means little without the reach behind it, and our guide to how a station proves its reach keeps those two apart.

Takwimu Intelligence (TI) holds that reach, appreciation and effect are separate measures. A campaign report should show each one, with its own margin, and never roll them into a single score.

Where TI fits

A broadcast lift study is survey work, and TI is built for surveys. Questions are asked of a sample, weighted to KNBS census totals at county level, and cross-tabbed by station listening, with the margin of error shown on each figure and open answers coded.

To see it on your own station or title, book a walk-through.

Takwimu Intelligence (TI) is audience measurement software for Kenyan media.

Sources

  1. 1

    About Brand Lift · Google Ads HelpOpened 5 November 2026.

  2. 2

    TikTok Brand Lift Studies · TikTok Ads HelpOpened 5 November 2026.

  3. 3

    Set up Brand Lift measurement · Display & Video 360 HelpControl group and responses needed. Opened 5 November 2026.

  4. 4

    Understand your Brand Lift measurement data · Display & Video 360 HelpMetric definitions. Opened 5 November 2026.

  5. 5

    A New Benchmark for Mechanical Avoidance of Radio Advertising · Michelon, Bellman, Faulkner, Cohen and Bruwer, Journal of Advertising Research, March 2020Full paper. Opened 5 November 2026.

  6. 6

    GeoPoll Audience Measurement · GeoPollMethod and tools, from its own page. Opened 5 November 2026.

More from the library