How to read a brand lift study in influencer marketing

The study answers one question, so pick it early

A brand lift study asks a group of people a question, asks a comparable group the same question, and reports the gap. That is the whole idea. Everything that makes it credible or useless happens in the setup.

So the first decision is which question. A product nobody has heard of needs awareness. An established brand that keeps losing the shortlist needs consideration. These are different surveys, and choosing between them after the campaign has run means choosing the one that looks best, which is not measurement.

Make the choice before the creative and the media plan are final, so the study measures the thing the campaign set out to move.

Keep survey answers and observed actions in separate columns. Someone telling a survey they intend to buy is not a purchase. Both belong in a report; naming them the same thing does not.

Who is in each group is the whole study

Two questions decide whether a lift number means anything, and neither is usually on the slide.

First: how was exposure determined? Second: how was the control group selected? Random assignment and a matched observational comparison are not the same evidence, and they do not support the same claim. A control group that looks similar on a few demographics has not automatically removed selection bias. People who saw a creator campaign about running shoes may differ from people who did not in exactly the way that matters.

Then ask which channels the study covers. A study run inside a paid platform sees what that platform served. A creator campaign that also ran organically, got reshared, and picked up a mention in a newsletter has exposure the study cannot see. When members of the control group have in fact seen the campaign, the gap between the groups shrinks and the study understates a real effect.

None of this makes a study worthless. It makes the study a measurement of something specific, and worth stating plainly in the report.

Percentage points and relative lift are different numbers

This is the one that causes arguments in meetings, and it is arithmetic.

Illustrative example, not a Beatly campaign result. In the exposed group, 30% give a positive answer. In the control group, 25% do. The absolute difference is 5 percentage points. The relative lift is 20%, because (30 − 25) ÷ 25 = 0.2.

Both describe the same result. One sounds four times more impressive. A report that says only “+20%” has told you the ratio and hidden the base, and a ratio on a small base can be enormous while the underlying movement is tiny.

Ask for both response rates, the calculation, and the uncertainty around the estimate. Any vendor doing serious work has all three ready.

Read the uncertainty before you pick a winner

Every survey has noise. Two groups of people who saw nothing at all will still differ by a few points, purely by chance. A study can only call a lift real when it is bigger than that chance gap, and that gap shrinks as the sample grows.

Which means a study can miss a real result. A campaign that genuinely moved awareness by five points, measured with too few responses, comes back as no significant difference. The campaign worked. The study could not see it.

So ask for the sample size in each group, the fieldwork dates, and the confidence interval. Then ask the question almost nobody asks: was this study designed to detect a difference of the size I care about? If the answer is vague, that is the answer. Our Lift Test does this arithmetic in front of you, before you pay for anything.

Be especially careful with subgroup results. “It really worked with women 25 to 34” is often a small slice of an already small sample, and small slices produce large-looking numbers.

An inconclusive result is a finding about what the study can support. It is not permission to write a more confident headline. Keep the limitations next to the number when the report gets forwarded, because it will get forwarded.

Put this in the brief, not in an email afterwards

Every decision above has to happen before launch. Sample sizes cannot be fixed retroactively, the control group cannot be chosen after the fact, and the survey question cannot be rewritten once the fieldwork is done.

Bring these to a Beatly strategist while the campaign is still a plan. The measurement scope has to fit the audience, the delivery and the budget, and that is a conversation, not a checkbox.

• State the primary survey outcome and target market.
• Ask for study eligibility, cost and channel coverage before launch.
• Agree on exposure, control selection and fieldwork timing.
• Request group sizes, response rates, uncertainty and limitations in the report.