Where do these numbers come from?

From the standard sample-size formula for comparing two proportions, at a two-sided significance level of 0.05 and 80% power. It is the same calculation any statistician would run, and you can check it against any statistics text. Nothing here uses Beatly data.

Why does a small campaign need such a big sample?

Because the smaller the effect you want to see, the more responses it takes to tell it apart from noise. Halving the effect you want to detect roughly quadruples the sample you need. A study sized for a 10 point lift is useless for a 3 point one.

What is a realistic lift to expect?

It depends on the category, the baseline, the media weight and the creative. Beatly does not publish a benchmark here because a number without its campaign attached would be misleading. Ask any vendor quoting a benchmark for the sample size, the market and the field window behind it.

What if the study comes back with no significant difference?

One of two things, and this page tells you which before you spend the money. Either the campaign genuinely did not move the metric, or the study was never big enough to see a lift of that size. Only the first is a finding.

Can I skip the control group?

You can measure before and after, but you cannot attribute the difference to the campaign. Seasonality, other media and a competitor going quiet all move the same numbers. The control group is what separates your effect from everything else that happened that month.

Free tool · Brand lift

Will it lift?

Before you pay for a brand lift study, check it could actually see your result. Small studies miss real wins and report “no difference”. Drag the three sliders. The yellow band shows how big a win has to be before this study would notice it.

The two numbers that belong to your survey partner

Share of the exposed audience reachable in panel %

Survey completion rate %

Defaults are the cautious end of what panel vendors quote. Ask yours for both and type the real ones in. Beatly does not pretend to know them.

Almost.

Right on the noise floor. One slow week and it reads as nothing. Push reach, narrow the market, or agree what a null result means before you buy the study.

Didn’t see the campaignSaw itToo close to call: a study this size can’t tell these apart from luck

If the green tower clears the yellow band, the study can prove the lift. If it stays inside, it can’t.

4.4ppSmallest win this study could notice

2,800People who’d actually answer the survey (1,400 per group)

312KPeople you’d need to reach to prove +5pp

Talk it through → Add this to a brief

What just happened

The floor is the whole story.

A lift study asks two comparable groups the same question. One saw the campaign, one didn’t. The gap between them is the lift.

But every survey has noise. With a small sample, two identical groups can differ by a few points just by chance. The study can only call a lift real when it is bigger than that chance gap. That gap is the noise floor, and it shrinks as the sample grows.

Halve the lift you want to see and you need roughly four times the answers. That is why “we ran a brand lift study” and “we could have seen the result” are two different sentences.

The maths, shown

Check it yourself.

Both tables use the two-proportion sample size formula at 0.05 significance and 80% power, the convention brand lift vendors report against. Nothing here is Beatly data.

Completed answers needed, per group
Awareness today	+2pp	+3pp	+5pp	+8pp	+10pp
10%	3,839	1,772	683	292	197
20%	6,507	2,940	1,091	444	291
30%	8,391	3,760	1,374	547	354
45%	9,745	4,336	1,562	610	389

Smallest lift a given sample can see
Awareness today	200 a side	400 a side	600 a side	1,000 a side	1,500 a side	2,000 a side
10%	9.9pp	6.7pp	5.4pp	4.1pp	3.3pp	2.8pp
20%	12.2pp	8.4pp	6.8pp	5.2pp	4.2pp	3.7pp
30%	13.4pp	9.4pp	7.6pp	5.9pp	4.8pp	4.1pp
45%	13.9pp	9.9pp	8.1pp	6.2pp	5.1pp	4.4pp

Read the second table as a warning. With 400 answers a side and awareness at 20%, anything under 8.4 points is invisible. A campaign that genuinely moved awareness by five would be reported as no difference.

Before you commission a study

Questions people ask us.

Where do these numbers come from?: From the standard sample-size formula for comparing two proportions, at a two-sided significance level of 0.05 and 80% power. It is the same calculation any statistician would run, and you can check it against any statistics text. Nothing here uses Beatly data.
Why does a small campaign need such a big sample?: Because the smaller the effect you want to see, the more responses it takes to tell it apart from noise. Halving the effect you want to detect roughly quadruples the sample you need. A study sized for a 10 point lift is useless for a 3 point one.
What is a realistic lift to expect?: It depends on the category, the baseline, the media weight and the creative. Beatly does not publish a benchmark here because a number without its campaign attached would be misleading. Ask any vendor quoting a benchmark for the sample size, the market and the field window behind it.
What if the study comes back with no significant difference?: One of two things, and this page tells you which before you spend the money. Either the campaign genuinely did not move the metric, or the study was never big enough to see a lift of that size. Only the first is a finding.
Can I skip the control group?: You can measure before and after, but you cannot attribute the difference to the campaign. Seasonality, other media and a competitor going quiet all move the same numbers. The control group is what separates your effect from everything else that happened that month.

Design the study in.

The brief decides the survey question. If you want to prove the lift, that decision has to happen before anybody films anything.

Talk to a strategist →