What Counts as True Incremental Revenue From Display Ads?
True incremental revenue from display ads is the share of sales that would not have happened without ad exposure. It excludes purchases from repeat buyers who were already planning to buy. It also excludes conversions credited to an ad only because a shopper saw it before purchasing. Standard attribution models report every conversion that falls within a set window after an ad view or click. They do so whether or not the shopper would have bought anyway. That shopper may have arrived through organic search, direct navigation, or loyalty to the brand. The result is a gap between reported conversions and actual incremental impact. Incrementality testing closes that gap by comparing an exposed group with a comparable unexposed control group. This comparison isolates the lift that the advertising itself produced. The difference between the two groups represents the true incremental effect, not the raw conversion count. New-to-brand metrics add to this picture by identifying purchases from customers who had not previously bought the product. These metrics offer a partial signal of acquisition spending versus retention spending. A sale attributed to a display ad is not automatically a sale caused by that display ad. Separating the two requires a baseline of expected sales. Performance is then measured against that baseline rather than against raw attributed totals. Campaigns that look efficient under standard attribution can still generate little or no incremental revenue. That becomes visible once cannibalized and organic-leaning sales are removed from the total. Measuring true incremental revenue means separating what the advertising created from what the business would have earned without ad spend.
Last Updated: October 26, 2026
- • What Incremental Revenue Actually Measures
- • Where the Line Between New and Repeat Blurs
- • Why Attribution Windows Miscount Display Ads Both Ways
- • How Attribution Models Get Built to Claim Credit
- • Running an Incrementality Test Without Guesswork
-
• Frequently Asked Questions
- • What is the difference between attribution and incrementality in Amazon ads?
- • How do you run an incrementality or lift test on Amazon?
- • What are new-to-brand metrics and how do they relate to incremental revenue?
- • Can you measure incremental revenue without advanced tools like Amazon Marketing Cloud?
- • How long should an attribution window be for display ad campaigns?
- • What This Means
What Incremental Revenue Actually Measures
True incremental revenue is the portion of sales that would not have happened at all if your display ads were turned off. That definition sounds simple. Getting to the real figure is part of what Amazon Ads Management does with display budgets. The default dashboard was never built to answer it.
It’s the first problem you hit when making sense of Amazon display ad profitability, and it shows up before you ever reach a control group.
There’s a lot riding on the answer. Amazon reported $21.3 billion in advertising services revenue in the fourth quarter of 2025, up 23 percent from the same quarter a year earlier. With that much money in play, it pays to know which part of a display budget actually created sales.
Where the Line Between New and Repeat Blurs
Default dashboards were built to count sales, not to sort them.
Telling new customers from loyal ones is the central challenge in measuring display campaigns, and nothing in standard reporting solves it on its own.
ACoS and view-through conversions are the easiest numbers to pull, so they get leaned on hardest, yet neither one shows who was new. Spotting genuine new buyers in ad data starts with separating those two buyer types before any optimization decision gets made.
Why Attribution Windows Miscount Display Ads Both Ways
Attribution windows are not neutral measurement tools. They are time boxes, and anything that happens outside the box gets excluded from the report regardless of whether the ad caused it.
The Two-Week Window Problem
Take a two-week attribution window. Inside it, different ad types finish showing their effect at very different speeds.
That’s not the window’s only flaw: repeat buyers inflate display campaign numbers too. Loyal customers keep buying on their own schedule, and when one of those purchases lands inside the window, it gets counted as an ad sale.
Why Upper-Funnel Ad Products Get Shortchanged by Design
Lower-funnel ad products sit right next to the purchase. A shopper clicks a sponsored product ad and buys within days, so the two-week window catches most of the effect.
Upper-funnel display exposure works differently. It plants awareness that a shopper acts on later, often after the window stops counting. An Amazon Science study found that within the two-week attribution window, upper- and middle-funnel ad products show only 30–50% of their total effects, compared with 60–90% for lower-funnel products. So the same window that over-credits repeat buyers under-credits the ads that reach new ones.
How Attribution Models Get Built to Claim Credit
An attribution model is built, choice by choice, to assign credit somewhere. Every design decision inside it decides which touchpoint gets that credit and which gets ignored.
Amazon’s own methodology puts those choices on display. Look at how it weighs touchpoints, and it’s clear why a single-touch report can’t answer the incrementality question by itself.
| Measurement Approach | What It Credits | What It Misses |
|---|---|---|
| Last-click attribution | The final touchpoint before purchase, regardless of what influenced the decision earlier | Upper-funnel exposure that built the intent a shopper acted on later |
| Single-window attributed conversion reporting | Any purchase inside the set time box after an ad view or click | Whether the sale came from the ad, organic search, or brand loyalty |
| Multi-touch attribution | A proportional share of credit across several touchpoints, weighted by likely contribution | Whether one campaign caused the sale on its own; Amazon’s version uses randomized trials alongside machine learning, but it still splits credit rather than measuring one campaign’s lift |
| View-through conversion tracking | A sale after an ad impression with no click at all | Confirmation that the impression had any effect on the shopper’s decision |
| Incrementality testing | The difference in outcomes between an exposed group and a matched unexposed control group | Which single touchpoint deserves credit, since it measures total lift instead |
Multi-Touch Attribution and Its Machine-Learned Weighting
A 2025 arXiv preprint describes Amazon’s Multi-Touch Attribution as using randomized controlled trials and machine learning models to allocate credit for Amazon conversions across Amazon Ads touchpoints in proportion to their likely contribution to shopping decisions. That is a meaningfully different architecture than last-click or single-window reporting.
Why ACoS-Only Measurement Fails as a Growth Signal
ACoS tracks spend against attributed sales inside a single campaign. It says nothing about whether the business grew, which is how a channel can post an efficient ACoS while total revenue sits flat. TACoS answers that question, because it measures ad spend against total sales across the whole account.
Running an Incrementality Test Without Guesswork
An incrementality test turns the attribution problem into a design problem. The question stops being which touchpoint gets credit. It becomes what happens to sales when the ad disappears entirely.
Here’s the idea in made-up numbers. Say 10,000 shoppers see the ads and a matched 10,000 don’t. If 300 people in the exposed group buy and 250 in the control group buy, the ads added about 50 sales, not 300. The other 250 were coming either way.
| Implementation Step | What It Isolates | Data Source |
|---|---|---|
| Define the control group | A baseline of what shoppers buy with no ad exposure | Holdout audience withheld from the display campaign, matched against the exposed segment |
| Hold the exposed and unexposed groups constant for the test window | Whether purchase differences reflect the ad or ordinary variation in buying behavior | Campaign delivery settings and audience assignment inside the ad platform |
| Compare purchase outcomes between groups | The lift the display ad added on top of that baseline | Purchase data drawn from both the exposed and control cohorts over the same window |
Where Amazon Marketing Cloud Fits Into the Test
Amazon Marketing Cloud lets an operator pull purchase outcomes for the test’s holdout and exposed groups, across touchpoints and funnel stages. The matched split built into the test is what makes that comparison a lift read; AMC is where the numbers come together.
Frequently Asked Questions
A few mechanical questions tend to surface once the measurement problem is clear. The answers below settle the remaining definitional gaps.
What is the difference between attribution and incrementality in Amazon ads?
Attribution assigns credit to a touchpoint inside a reporting window. Incrementality asks a harder question: would the sale have happened without that touchpoint at all?
How do you run an incrementality or lift test on Amazon?
A lift test withholds ad exposure from a matched control group while an exposed group sees the campaign. The gap between the two groups’ purchase rates is the incremental result, not the exposed group’s raw total.
What are new-to-brand metrics and how do they relate to incremental revenue?
New-to-brand metrics flag purchases from shoppers who haven’t bought from the brand within a set lookback window. That’s a useful acquisition signal, but not proof the ad caused the sale. Pair it with a lift test to see what the ad actually added.
Can you measure incremental revenue without advanced tools like Amazon Marketing Cloud?
A simpler lift test can be built by comparing matched markets or time periods with and without display spend. It’s rougher than a holdout on matched audiences, because markets and time periods differ in ways the test can’t control, but it still separates new revenue from cannibalized revenue better than attribution alone.
How long should an attribution window be for display ad campaigns?
Longer than two weeks for upper-funnel display. An Amazon Science study found upper- and middle-funnel ads showed only 30–50% of their effect inside a two-week window, so judge them on a longer read than lower-funnel ads sitting next to the purchase.
What This Means
So treat the next display report as a starting claim, not a verdict. Before any test begins, write down a baseline: weekly total sales and TACoS for the account.
Then run one lift test on a single display campaign with a matched control group, and keep the read open longer than two weeks.
That’s the standard behind Amazon DSP management: display spend measured against total revenue movement, not attributed sales alone. Start with a free Amazon audit.