Brand lift: measuring recall against a baseline

A holdout that sees nothing gives you a baseline. Lift is the distance your creative puts between them, read for exactly what it is worth.

Most tests ask which of two versions landed better. A brand lift study asks something else: did this piece move anything at all?

The holdout

Half the people recruited watch your creative. The other half watch nothing and answer the identical questions. That second group is the holdout, and their answers are your baseline.

Ask a room of strangers which brands they have seen advertising for lately and some will pick yours whether or not they saw anything. A few genuinely remember something, a few are guessing, a few know you from somewhere else entirely. That rate is the bar. Lift is the distance between the two groups: what the exposed group reports, minus what the holdout reports.

Why the control answers the same questions

The holdout takes the same survey, in the same words, in the same order. Exposure is the only difference between the groups, which is what lets you read the gap as a consequence of exposure. Ask the control something different and there is nothing to subtract.

What the design nets out, and what it does not

The holdout measures guessing and the awareness people already carried in the door, and lift takes both out of your number. What it does not remove is the effect of having watched something at all: the control sees nothing, so it is not a matched placebo. Read the result as lift over a no-exposure baseline. It is never proof that the creative alone caused the change.

Picking the other brands on the list

The recall question lists your brand among other brands, with a "None of these" option that cannot be removed. Pick real brands your audience could plausibly have seen: competitors, or names from an adjacent category. Invented brands flatter you, because nobody recalls something that does not exist, so the baseline sinks and your lift looks larger than it is. "None of these" stays for the same reason: forcing a choice inflates both groups at once.

Panel only, and why a share link cannot be a baseline

Brand lift recruits a panel and refuses a share link. Your own list is self-selected: they opened your email, they follow the work, many of them already know the brand. A holdout drawn from those people would report high recall having seen nothing, which does not make your creative worse, it makes the comparison meaningless. Refusing that is part of the design rather than a missing feature.

When lift is not confirmed

Sometimes a study cannot confirm a lift. That is informative, and it still ships a number worth having: the baseline. Knowing what share of your audience claims recall of you before any exposure tells you the height of the bar, which is the effect your next piece has to produce to be visible at all.

When to reach for it

Use brand lift for one finished piece with real spend behind it, when the question is whether it moved recall or favorability at all. If you are choosing between two cuts, run a normal two-version test instead: this design measures one creative against nothing, not one cut against another. See Choosing a test type.