Brand Lift

Performance advertising reports itself. Put $3,200 into ads, get 80 signups, and you can say each one cost $40. Brand advertising can't produce that sentence. Almost nobody signs up while watching a video ad, and when someone arrives through search two months later, there's no way to know whether they saw it.

So the same conversation kept repeating at Pagedeck. Grace would ask what $30,000 of brand campaign had bought, and Ravi would bring back 400,000 views and a comment count. That's a measure of how much advertising went out, not how much changed in anyone's head. With no answer available, the next quarter's budget drifted toward the side that could be measured.

Asking people about perception directly is an obvious next thought. The trouble is that surveying once before the campaign and once after doesn't answer it either. An article may have run in between, or a competitor may have gone quiet, so there's no way to tell whose the six-point rise was.

Brand lift solves that by manufacturing something to compare against. Over the same period, the same questions go to people who saw the ads and people who didn't, and only the gap between the two answers counts. The article and the season hit both groups equally, so what's left belongs to the advertising.

Subtract one group's response rate from the other's. The arithmetic is one subtraction; the actual work is building the two groups.

1. Before the campaign runs, split the audience at random. One side sees the ads; the other sees an unrelated ad in the same slot, or nothing. Randomizing is the same device an A/B test uses. It spreads age, region and everything else evenly across both sides, so any gap that shows up later can't be explained by something other than the ads.
2. When the campaign ends, ask both groups the same question at the same time. Something like "select every collaborative document tool you're aware of." The wording has to be identical down to the character.
3. Subtract the two rates. If 34 percent of the exposed group picked Pagedeck against 26 percent of the control group, the lift is 8 percentage points.
4. Write the relative figure too: 8 divided by 26 is 31 percent. The same 8 points reads as a bigger jump when the starting number is low, so both values belong in the report.

What you ask determines which layer you're measuring. Give people the name and ask whether they know it and you're measuring brand awareness. Ask what comes to mind for "collaborative document tool" with no name given and you're measuring brand recall. Questions about intent to use or favorability often ride along. One study usually carries all of them, and each layer gets read separately.

Sample size is where this gets expensive. At 400 people per group the margin of error runs around ±5 percentage points, which cannot resolve an 8-point difference. Seeing a gap that size takes thousands per group, which is why lift studies are usually run for you by the large ad platforms.

An outside benchmark is no use here. Change the wording slightly and the response rate changes wholesale, so someone else's 10 points and your 8 points are not comparable numbers. The standard is your own last campaign, measured with the same question in the same way.

Expectations differ by layer. Awareness, which only asks whether people recognize the name, moves several points when enough media runs behind it. Intent to use barely moves, so 1 or 2 points is a strong result. A campaign that produces 8 points of awareness and zero on intent isn't a failure; that's the usual shape.

Anything inside the margin of error reads as unknown, not as zero. A 2-point lift with a ±3-point margin doesn't mean the advertising did nothing, it means this study can't tell. Conflating the two gets healthy campaigns cancelled.

To compare campaigns, attach the cost. Spending $30,000 to raise awareness by 8 points puts each point at $3,750, and if last quarter's point cost $6,000, this creative was the cheaper one. The other thing worth watching is persistence. Re-run the same survey four weeks after the campaign ends and the raised number has usually come back down. How much of it stayed is a real input into which creative to run again.

"This campaign came back at 8 points of lift on awareness. Intent didn't move."

The layers split, so the next creative decision should target one of them deliberately.

"A before-and-after comparison with no control group doesn't get called lift."

The number failed to rule out causes other than the advertising.

"At 400 per group the margin of error is wider than the lift. Let's skip the study this time."

The measurement can't produce an answer under those conditions, so it's only a cost.

When Pagedeck closed out its quarterly brand campaign, the table Ravi carried into the review had three rows: awareness up 8 points, recall up 2, intent to use flat. His reading was that the name had spread but hadn't made it as far as being a candidate anyone lists.

Grace asked less about the numbers than about the next decision. Last quarter's video left the viewer with a name and a mood and never said what the product was; the next one will put the category and the name on screen together. Recall only rises when people remember which shelf the name belongs on.

The most common mistake is reading lift as revenue. It's a percentage and it came out of an advertising report, so the leap is easy to make. But the number is a difference in how many people raised a hand in a survey. Between saying you know a name and opening a wallet sit several months and several decisions. Plan a budget on that conflation, and an 8-point awareness lift booked as 8 percent more revenue falls apart at quarter end, usually landing on sales as if it were their miss.

Reading a zero as proof the advertising failed is the other one. A small sample hides a real effect, and so does a question phrased around a category your product doesn't sit in. Check the measurement conditions first, and only blame the creative when those conditions hold up.