How brand lift studies measure awareness, and the minimum budgets the platforms set

A brand lift study is one way advertisers try to measure awareness. The platform shows an ad to one group, holds it back from a group that was eligible to see it, and surveys both.

The price of entry, the size of the effects it can detect and the way results are expressed all matter to an executive deciding whether to commission one. This explainer sets out what Google and LinkedIn document about their studies.

What a brand lift study compares

In Google's Display & Video 360 documentation, a study compares a sample group, made up of people shown the ads, with a control group of people who were eligible for the ads but were not shown them. It measures changes in ad recall, brand consideration and brand awareness.

LinkedIn describes the same structure. For the length of the test it splits the target audience into a test group that can see the ads and a control group that cannot. Surveys go to a subset of active members in both groups, and the difference in answers is read as the ads' effect on perception of the brand. LinkedIn also says it estimates the return on the investment.

Google Ads lets an advertiser measure up to three survey questions, with metrics including ad recall, awareness, association, consideration, favorability, purchase intent or a custom metric. The study applies to video, audio or Demand Gen campaigns. For Demand Gen, Google says measurement won't capture results from other channels such as Gmail or Discover, and advises prioritizing YouTube video ads.

What the minimum budgets are

Google ties the minimum to the number of survey questions and to the country. For one question, the minimum is $5,000 in its lowest group of countries (Country A), $10,000 in Country B and $15,000 in Country C. Two questions need $10,000, $20,000 and $30,000. Three questions need $20,000 in Country A and $60,000 in both Country B and Country C.

The United States and United Kingdom sit in Country B. Canada, Germany, Japan and Australia are among the Country C markets. Brazil, India and Mexico are among those in Country A. A study that targets countries in more than one group takes the highest minimum, and countries not listed default to the highest group.

Eligibility is judged on the total budget of the active campaigns being measured, counted over the 10 days from the study's start date. Paused campaigns do not count. Where a campaign has a total budget instead of a daily one, Google divides it by the number of days, so extending the end date can make a study ineligible.

Access is not automatic. Google says Brand Lift is not available to all Google Ads accounts, and advertisers must contact their account representative.

How long a study runs and how many people it needs

A standard Google Ads study completes in 3 to 14 days. An Enhanced Lift study runs 9 to 14 days and needs three times the minimum budget, to collect about three times as many survey responses. Surveys take up to 48 hours to be approved and then run automatically once the campaigns start.

In Display & Video 360, studies on YouTube and partner inventory run for 14 days or until enough responses are collected. Studies across other exchanges, such as connected TV, are measured as a snapshot and run for 40 days or until enough responses arrive. Connected TV has no required budget, but Google highly recommends at least $15,000 per survey question.

Response volume sets the smallest effect a study can pick up. Google's tables list 1,200 to 2,800 responses for detecting an absolute lift above 4%, and 5,000 to 11,000 for 2% in Display & Video 360 (4,100 to 11,000 in Google Ads). Detecting 1% takes 20,000 to 45,000 responses, and under 0.5% takes more than 180,000.

Results may appear at about 2,000 responses per lift metric for high-performing line items, or about 4,100 at the recommended minimum budget. If no lift appears after 16,800 responses, Google says lift may not be detectable.

In LinkedIn's Brand Lift Test API documentation, a test needs at least two survey questions, with recall mandatory. Its tests cannot run for more than 90 days, and results can keep changing for up to 15 days after a test ends.

How lift is reported

Google reports several lift measures, including absolute lift and relative lift. Absolute lift is the difference between the positive response rate of the exposed group and that of the baseline group, the people who were held back from seeing the ads. Moving from 20% to 40% is a 20-point absolute lift.

Relative lift divides that difference by the baseline group's positive responses, so the same move from 20% to 40% is a 100% relative lift. Because it is calculated from sampled responses, Google presents it as an estimated range with lower and upper bounds.

On the confidence level, Google's help page gives two figures. One passage says lift results use 80% two-sided confidence intervals, and another says the interval is usually 90%. Google's example of a 35% relative lift with a range of 30% to 40% describes a 90% chance that the true lift is above 30%.

A study can show "Not enough data," when responses fall below the minimum threshold, or "No lift detected," when there are enough responses but no statistically significant gap between the groups.

For a cost figure, Google calculates cost per lifted user as total campaign cost divided by the number of lifted users.

Where studies do not run or do not reach

Google lists several exclusions for cross-exchange studies. Connected TV measurement is not available in the EU, Netflix inventory is not measurable and digital out-of-home inventory is not supported. Alcohol, gambling, health, pharmacy and politics topics are not supported, and a cross-exchange campaign that mostly targets web inventory may not get enough survey responses.

Survey exposure is capped. A viewer gets at most one survey a day and three a week, can respond once per product or brand within 14 days, and sees one randomly chosen question. Only aggregated answers are shared.

Data slices appear in Display & Video 360 only with at least 250 responses. Google adds that advertisers may notice delivery and performance fluctuations when cookies or device IDs are unavailable.

A study also cannot normally be extended once its duration is set. Google's stated options are a new flight or a re-measurement, and a study can be ended early once it reaches a response threshold.