How we measure brand lift

Three study designs

We match the design to the signals the media environment has.

• Deterministic Brand Lift. For environments with deterministic ad serving. Respondents are assigned to exposed or control groups by the ad server and surveyed in-flight. This is our standard design and gives causal lift.
• Baseline Brand Lift. For environments with limited ad technology. Pre and post surveys around the campaign flight measure market movement. Results are reported as Directional by design.
• Adaptive Brand Lift. For CTV, choice-driven formats and privacy-first environments. We use the strongest design the environment supports and state the confidence level of every result.
• Out-of-home, cinema and AI answers For out-of-home and cinema we build exposed and control groups from where people travel and whether they went to the cinema during the campaign. For AI answers we combine visibility tracking with survey reads among real category shoppers. Both are Adaptive designs, so every result states its confidence level.

The five KPIs

Lift for each KPI is the exposed response rate minus the control response rate, in percentage points.

Brand Lift Score

The Brand Lift Score is 50 plus the average lift across the five KPIs, on a 0 to 100 scale. 50 represents the European digital benchmark (indicative), aggregated from published industry norms (Meta, Google/YouTube, Kantar, Nielsen, Dynata). It is a reference point, not a category standard.

Lifted users and eCPLU

• Brand Lifted Users = campaign reach × consideration lift. This estimates how many reached users moved into active consideration because of the campaign.
• eCPLU (effective cost per lifted user) = total media spend ÷ Brand Lifted Users. It measures media efficiency, not the cost of the study.

Where consideration isn't measured, we use Favorability.

Significant or Directional

We test every KPI lift for statistical significance at 95% confidence, comparing exposed and control groups.

• Significant the lift is unlikely to be down to chance.
• Directional the lift is positive but rests on a smaller base. It is an indication, not a confirmed result.

Every KPI in every report carries one of these two labels.

Sample sizes

Precision depends on the number of responses in each group, not the total. As a guide, with about 400 responses per group a lift can be measured to within about ±7 percentage points. Platform studies built to detect 1-point lifts need tens of thousands of responses. We size studies to detect the lifts that matter for a media decision, typically several points, and we say so up front.

Quality checks

Before analysis we remove speeders, duplicate responses and retargeting bleed (respondents shown the campaign outside the measured flight). An analyst reviews every study, including open-ended answers, before the report is released.

Privacy and data handling

• No cookies and no device IDs are used to measure exposure.
• Surveys collect no names, email addresses or other direct identifiers.
• Survey data is processed and stored on EU infrastructure.
• Results are reported in aggregate only. We never report on individuals.
• Data processing agreements are available on request for agency and publisher partners.