How to Measure Brand Lift From Your Advertising Campaigns

Brand campaigns get judged unfairly all the time. A YouTube and OTT launch runs for a month, someone pulls the click-through rate, and the campaign is declared a flop. Clicks were never the job. The job was to make more people remember, know and consider the brand, and that's what brand lift measures.

What is brand lift?

Brand lift is the change in how people think about your brand that is caused by your advertising. It's measured by comparing two groups: people who saw your ads (the exposed group) and similar people who were held back from seeing them (the control or baseline group). Both groups answer the same short survey question, and the difference in positive answers is the lift.

Because the two groups are split at random before the campaign reaches them, the difference can be put down to the ads rather than to a festive season, a competitor's price cut or anything else happening in the market.

The usual metrics line up with the stages of the funnel:

How do you read brand lift results?

Google's Brand Lift, available in DV360 and Google Ads, runs this experiment for you. Eligible users are randomly split, some are held back from your ads, and short survey questions go to people in both groups. Results come back as a handful of numbers worth understanding before you open a report:

• Absolute lift: the exposed group's positive response rate minus the baseline group's rate.
• Relative lift: absolute lift divided by the baseline rate.
• Headroom lift: absolute lift divided by the share of the baseline group that didn't answer positively. It shows how much of the available room for growth your ads captured.
• Lifted users: Google's estimate of how many people changed their answer because of your ads.
• Cost per lifted user: campaign cost divided by lifted users.

Here's an illustrative example. A spice brand launches a new range in Tamil Nadu and Karnataka with YouTube and connected TV and OTT, spending ₹30 lakh over four weeks and measuring awareness and ad recall. In the awareness question, 30% of the baseline group name the brand, against 36% of the exposed group.

• Absolute lift = 36% - 30% = 6 percentage points
• Relative lift = 6 / 30 = 20%
• Headroom lift = 6 / (100 - 30) = 8.6%
• If the study estimates 4,00,000 lifted users, cost per lifted user = ₹30,00,000 / 4,00,000 = ₹7.50

Google reports these as estimates with a confidence interval, usually 90%, so treat small differences between two studies with care. Cost per lifted user is the figure to carry forward. Next time you can test a different creative, channel mix or frequency and see which one buys more change per rupee. It is a useful yardstick for FMCG brands deciding where launch money goes.

How do you set up Brand Lift in DV360?

1. Check the inventory is eligible. Brand Lift in DV360 covers YouTube & partners video and audio line items, including instant deals, and cross-exchange line items for connected TV, display, video and audio bought through real-time bidding or instant reserve. DOOH can't be measured, and non-YouTube Programmatic Guaranteed inventory is exempt from measurement. Premium OTT is often bought on PG deals, so check eligibility before you promise anyone a lift study.
2. Check the budget. For YouTube & partners, Google sets minimum budgets over a 10-day window by country group and number of questions. Google's published table puts India in the lowest group: US$5,000 for one question, US$10,000 for two and US$20,000 for three, converted to your account currency. Cross-exchange CTV has no fixed minimum, but Google recommends at least US$15,000 per question. These figures change, so trust the eligibility check you see when you create the study.
3. Create the study. At advertiser level, go to Experiments, open the lift studies tab and create a new study. Pick up to three metrics, then enter your brand or product name and up to three competitors.
4. Get the timing right. For cross-exchange inventory, Google recommends setting the study up and getting it approved before the insertion order starts; surveys then run for up to 40 days. For YouTube & partners, the study starts as the line items serve, and surveys run for up to 14 days or until enough responses come in.
5. Leave the budget alone. Any drop in budget during the measurement period can switch the study off, and removing an insertion order from it can affect eligibility. If a study is deactivated, Google's suggested fix is to raise the budget on the original campaigns.
6. Read results when they're ready. Strong line items can show lift at around 2,000 responses per metric. At the recommended budget, expect about 4,100. If nothing shows after 16,800 responses, the lift is probably too small to detect.

If part of your goal is to get people searching for you, DV360 also offers Search Lift on YouTube & partners insertion orders. It measures the increase in searches for your brand among people who saw your ads, compared with a control group. Its budget minimum matches a one-question Brand Lift study, and running both doesn't add the budgets together. Our guide to YouTube advertising formats covers the ad types that usually carry these studies.

How many questions should you measure?

As few as answer your real question. Each extra question raises the budget threshold and spreads the responses thinner. For most launches, ad recall plus awareness is plenty. For established brands, consideration alone often tells you more than three metrics with weak readings.

How can you measure brand lift without a Google study?

Plenty of campaigns fall below the threshold or run on inventory the study can't cover. You still have options, each with trade-offs.

• Geo split test. Run the campaign in one set of cities and hold it back in a matched set. Survey both before and after with a research panel and compare the change. The difference between the two changes is your lift. It costs more in research fees but gives the cleanest answer outside a platform study.
• Pre and post tracker. Survey your target audience before launch and again after. It's cheaper but weaker, because anything else that happened in between, such as a competitor's sale or a big cricket series, gets mixed in.
• Search and site signals. Compare branded search interest (Google Trends is fine for a rough read) and direct site traffic between test and control regions. These are signals rather than proof, but they cost nothing.

Whichever method you use, keep the control group clean and ask exactly the same question before and after.

What makes a brand lift study fail?

• Too little reach or frequency. If too few people saw the ad enough times, there's nothing to detect. Plan this first; our reach and frequency calculator covers the maths.
• Ads that weren't seen. Low viewability dilutes the exposed group. See how to improve ad viewability.
• A contaminated control group. If the same creative is running heavily on TV or other channels, the control group has seen it too, which can hide the lift. Targeting only people who already watched your videos has a similar problem, because the control group has seen your ads before too.
• Mid-study changes. New creatives, budget cuts or audience changes muddy the comparison.
• Too many questions for the budget.
• Reading "not detectable" as "failed". It can mean the sample was too small to see a change, not that the ads did nothing.

How does brand lift fit with ROI?

Brand lift is a middle measure. It shows that people's minds changed, not that sales followed. Use it alongside incremental reach for the top of the funnel, and the methods in our guide to measuring programmatic advertising ROI for the commercial result. If you're still shaping the campaign itself, start with how to create a digital branding campaign.

Next step

Brand Lift works best when it's planned in from the start, not bolted on after launch. Ask for a free media plan, or bring your own plan to us as a Campaign-Based Engagement and we'll check Google's eligibility for your channels and budget before anything goes live.