You measure brand lift by splitting an audience into people eligible to see your ads and people who are not, surveying both, and subtracting one result from the other. If you are comparing survey scores before and after a campaign, you are measuring the calendar, not your ads.
That is the whole idea. The rest of this guide is the detail that decides whether your number is real: which metric to pick, how many completed surveys you need, where contamination creeps in, and what to do when the result is small. We run paid media for e-commerce and DTC brands, so we have a stake in this: a brand lift number that cannot be trusted is a budget argument you lose.
Key takeaways
- Brand lift is the gap between an exposed group and a control group. Before-and-after surveys without a control do not isolate your campaign.
- A common vendor framework uses 200 exposed and 200 control completes (400 in total) to detect shifts of 3 to 10 percentage points at 90% confidence (Dynata). Smaller shifts are directional only.
- Pick one primary metric before launch. Awareness, consideration and purchase intent move at different speeds and for different reasons.
- Cross-publisher campaigns need a cross-publisher study. Per-platform studies can contaminate each other.
- Nobody publishes a universal good lift percentage. Judge your result against your own baseline, not a number from a blog.
Table of Contents
- What brand lift is, and what a study actually measures
- Why before-and-after surveys fail
- How to measure brand lift, step by step
- What is a good brand lift percentage?
- Measuring brand awareness without a full study
- Common mistakes that wreck a brand lift study
- Where brand lift fits next to ROAS and MER
- Our perspective: lift is a decision tool, not a trophy
- Getting the measurement and the media under one roof
- FAQ
- Sources
What brand lift is, and what a study actually measures
Brand lift is the increase in a brand metric caused by advertising, over what would have happened without it. The metric is usually awareness, favourability, consideration or purchase intent. The word that matters is caused.
A brand lift study is the research method that estimates that cause. The mechanism is a controlled experiment: people are randomly assigned to a test group and a control group, the test group is exposed to the media and the control group is not, and the results are compared to quantify impact (IAB Australia, Connected Television Handbook, 2020). Random assignment is what makes the comparison fair. Without it, the people who saw your ad might simply be the people who already liked you.
Snapchat describes its version plainly: the audience is randomised into a group eligible to see the ads and a group that is not, both groups get an in-app survey, and the difference in responses is the lift (Snapchat Business Help). Other platforms run the same logic with different plumbing.
Why before-and-after surveys fail
The most common unnamed habit in brand measurement is the pre/post tracker with no control. Run a survey in week zero, run it again in week eight, call the difference your campaign. It is not.
Here is why. Between those two surveys, a competitor ran a sale, the season changed, your PR landed or flopped, and organic word of mouth moved. All of it shows up in your post number. A control group experiences all of those things too, and does not see your ads. Subtract one from the other and what is left is your media.
Caveat: pre/post designs still have a use. A 2023 Australian field experiment used a between-subjects design with pre- and post-test questionnaires, and 1,075 participants completed the post-test measures, including sponsor-brand awareness (Monash University). Note the design: different groups, compared on the same measures. That is a controlled comparison, not a single tracker run twice.
How to measure brand lift, step by step
1. Choose one primary metric before launch
The IAB Australia Video Measurement Framework lists the brand effectiveness metrics: awareness, favourability or opinion, brand attributes or associations, consideration or preference, purchase intent, search or visit intent, likelihood to recommend, message association, brand fit or relevance, and ad creative diagnostics (IAB Australia, 2025). That is a menu, not an order form.
Pick one as the decision metric. Why: every extra metric is another comparison, and every comparison is another chance to find a fluke and call it a win. If you test ten metrics at 90% confidence, expect roughly one to look significant by chance alone. (More metrics ≠ more insight.)
A rough match between goal and metric:
| Campaign goal | Primary metric | Why it fits |
|---|---|---|
| Launching a new brand or product | Awareness | Low baseline, so there is room to move |
| Entering a crowded category | Consideration | Awareness exists, shortlist placement does not |
| Sales-led DTC push | Purchase intent | Closest survey metric to the till |
| Reputation or creative repositioning | Favourability or message association | Tests whether the idea landed |
The same framework says campaigns should be assessed through delivery, brand and outcome metrics together (IAB Australia, 2025). Brand lift is the middle layer. Impressions sit under it and revenue sits over it. If you are still deciding whether the business is ready to spend on media at all, a readiness check for paid advertising is a sensible first read.
2. Set up exposed and control groups that stay separate
The control group must never see the campaign. That sounds obvious until you run the same creative on Meta, YouTube and connected TV at once.
IAB Australia flags the problem for multi-publisher campaigns: a person can sit in the exposed group for one publisher and the control group for another, which muddies the read of the campaign’s overall impact (IAB Australia, 2020). Kantar Australia draws the practical line: publisher-specific lift can be enough for a single-publisher campaign, but a cross-publisher campaign needs an integrated cross-publisher study for an overall read (Kantar Australia).
So if you run on four platforms and get four platform lift reports, do not add them up. Each platform measures its own exposed and control groups, and the same person can be counted in several. Treat each report as a read on that platform, then use a single cross-channel study if you need one number for the board.
Another point on exposure. Dynata says its methodology does not force exposure: it tracks ad exposure through tags or platform integrations in natural media environments, to mirror real-world attention levels (Dynata). Why that matters: forced exposure tells you whether the ad can work. Natural exposure tells you whether it does work, when people scroll past it with a thumb.
3. Size the sample before you spend
This is where most small studies die. Dynata’s framework uses 200 exposed and 200 control respondents, 400 completes, to detect brand lift shifts of 3 to 10 percentage points with 90% confidence, balancing feasibility, cost and speed (Dynata).
Sit with that range. A 3-point shift and a 10-point shift are very different campaigns, and a study at that size cannot tell you which one you got when the true effect is small.
The maths is simple. With a baseline around 20% and 200 people per group, the noise on each group’s number is about 3 percentage points. The noise on the difference is about 4 points. At 90% confidence, that is a margin of roughly plus or minus 6 to 7 points. (Our own back-of-envelope arithmetic, not a vendor figure.) A true 3-point lift will hide inside that margin. Want to see smaller effects? You need more completes, and that costs more.
Dynata also says smaller lift shifts may remain directionally valid even when they are not statistically significant (Dynata). Fair. Just call them directional in the report, and do not put them in a headline.
4. Fix the flight length and the holdout
The holdout must be protected for the whole flight. If you pause the control for a week while a promotion runs, the contaminated week pollutes the result. Decide the flight length up front, because a short flight with a low frequency may simply not move a survey metric, and you will have paid for a study that could not have found anything.
There is an honest cost here. IAB Australia notes that controlled experiments are an effective way to measure incrementality but need time and resources, plus the opportunity cost of withholding exposure from a control group (IAB Australia, 2019). If you hold out 10% of your audience, you give up reach on that 10%. For most brands that is a fair price. For a brand with a tiny addressable audience, it can be a real one.
5. Read the result in the right units
Google defines absolute lift as the exposed group’s positive response rate minus the baseline group’s rate. It measures how much the ads shifted positive feeling toward the brand or product (Google Ads Help).
An illustration, with made-up numbers: baseline awareness is 20%, exposed awareness is 24%. Absolute lift is 4 percentage points. Relative lift is 4 divided by 20, so 20%. Both are true. Only one belongs in a sentence where you also say what it cost. Report both and label them.
Google Ads reports a positive lift when results are statistically significant, provides confidence intervals, and says results can offer directional insight at 50% certainty or higher (Google Ads Help). Read that last part carefully. Fifty per cent certainty is a coin flip with a lean. It is a reason to keep testing, not a reason to rewrite your media plan.
What is a good brand lift percentage?
Nobody has published a number that holds across categories, brands and platforms. Anyone quoting a universal good lift percentage is guessing, and the same goes for the 3-7-27 style rules of thumb that circulate in branding circles. We could not trace them to anything we would put in front of a CFO.
What does exist is a mechanism you can reason about. Baseline awareness caps how far awareness can move. A new DTC brand starting at 5% awareness has room to double. A household name at 70% does not. Comparing absolute lift across campaigns without accounting for different baselines leads to wrong conclusions, because newer brands have more room to grow (MediaGuides).
So benchmark against yourself. Your last study, same metric, same audience definition, similar flight length. That is the only comparison you can defend.
Measuring brand awareness without a full study
Not every brand can fund a 400-complete study per channel. If you are a founder spending a modest monthly budget, a lighter stack is better than none:
- Platform lift tools. Snapchat and Google run them natively, using the same exposed versus control logic. Check each platform’s minimum spend and audience requirements on its own help pages, because they change.
- Branded search volume. It is a behavioural proxy, not a lift measure. It moves for many reasons, so read it alongside a controlled test, never instead of one.
- Geo or audience holdouts. Switch off a region or segment and compare. Our write-up on incrementality testing for Meta ads covers the logic for performance channels, and the same discipline applies to brand.
- Your own post-purchase survey. Ask new customers how they first heard of you. Cheap, biased towards people who bought, and still useful as a sanity check.
A caution on all of these: survey-based measurement is sensitive to question wording, so keep the wording identical between waves and between groups.
Common mistakes that wreck a brand lift study
- No control group. Covered above. It is the big one.
- Too many metrics. Ten metrics, ten chances for a fluke.
- Underpowered samples. Fifty completes per group cannot detect a 3-point move. You will report noise as a finding.
- Summing platform reports. Overlapping audiences get counted more than once.
- Judging a short flight on a long-term metric. Consideration builds slowly. Test it after the media has had time to land.
- Changing creative mid-study without logging it. You then do not know what you measured.
- Treating a non-significant result as zero. It is unknown, not nil. Say so.
If your brand lift process lives across spreadsheets and email threads, tidy the marketing campaign workflow first. Study set-up errors are mostly handoff errors.
Where brand lift fits next to ROAS and MER
Brand lift answers a narrow question: did people think or feel differently because of the ads? It does not tell you whether the campaign paid back. For that you need outcome metrics and a budget model.
Our view is that the three layers should be read in order. Delivery tells you whether the media ran. Brand lift tells you whether it moved minds. Outcomes tell you whether it moved revenue. When the layers disagree, that is the interesting bit. Strong lift and flat sales usually means a conversion problem after the click. Flat lift and rising sales usually means performance media is harvesting demand that was already there.
For the outcome layer, see our workflow on MER versus ROAS, and for the longer-term budget picture, media mix modelling and when it pays off.
Our perspective: lift is a decision tool, not a trophy
Most brand lift reports we see get used as proof after the fact. A slide that says awareness rose, with no control group behind it and no decision attached. That is a vanity metric in a nice suit.
We use lift to decide something before the money moves. Which creative route earns a bigger budget? Does a new channel get a second flight? Does the brand message register at all, or are we paying for impressions that bounce off? A study that cannot change a decision is not worth fielding.
Three rules we hold to. First, one primary metric, written down before launch. Second, a result we cannot separate from noise gets reported as directional and does not get a victory lap. Third, lift never travels alone: it sits next to delivery and outcome numbers, because a brand that people recall but do not buy is a cost centre.
We are also blunt about the limit. Survey lift is a read on attitudes, not revenue. If someone promises you a clean formula from lift to sales, they are selling you a story. The honest version is slower and less tidy, and it is the one that survives a budget review.
Happy to share the study brief template we use if it would help.
— Harry
Getting the measurement and the media under one roof
The usual failure point is the handoff. One team buys the media, another makes the creative, a third runs the survey, and nobody owns the question the study was meant to answer. We run strategy, creative production, paid media and conversion work as one team, so the test design is agreed before the first ad goes live.
If you are planning a campaign across TV, streaming, radio or out-of-home alongside social, our integrated campaigns service is built for exactly the cross-channel problem described above. For a wider look at what we do for DTC and consumer brands, start at our marketing services page, or read how we worked with Sorted Drinks.
FAQ
How do you measure brand lift?
Randomly split your audience into an exposed group and a control group that cannot see the campaign. Survey both on the same metric, then subtract the control result from the exposed result. That difference, tested for statistical significance, is your lift.
How many survey responses do you need for a brand lift study?
One published vendor framework uses 200 exposed and 200 control completes to detect shifts of 3 to 10 percentage points at 90% confidence (Dynata). Want to detect smaller shifts? You need more completes. Treat anything below the detectable range as directional.
What is a good brand lift percentage?
There is no universal figure. Lift depends on your baseline, your category, your flight length and your metric, and newer brands have more room to grow than established ones. Compare against your own previous studies on the same metric.
Can you measure brand awareness across several platforms at once?
Yes, but not by adding platform reports together. The same person can sit in the exposed group on one publisher and the control group on another, which muddies the overall read (IAB Australia). Use one integrated cross-publisher study for the combined number.
Sources
- Snapchat Business Help: About Brand Lift
- Dynata+ Brand Lift FAQs
- IAB Australia: Connected Television Handbook 2020
- IAB Australia: A Guide to Designing Digital Ad Impact Studies, 2019
- IAB Australia: Video Measurement Framework 2025
- Google Ads Help: Lift measurement statuses and metrics
- Kantar Australia: why brand marketing campaign metrics are misleading
- Monash University: 2023 Australian field experiment
- MediaGuides: The Power of Brand Uplift Measurement
