DoorDash Ghost Ads: How a Platform Measures What an Ad Caused
%20(21).jpg)
Published:
October 20, 2025
Updated:
October 15, 2026
In September 2026 the Alliance for Audited Media certified DoorDash’s incrementality measurement for Sponsored Products and Sponsored Brands, the ads consumer brands buy on the app. In March 2026 DoorDash had launched Sales Lift Measurement for those brands, a test “based on Ghost Ads” that shows how many extra sales an ad produced. DoorDash said its Sponsored Products tests found “over 30% average sales lift for CPG advertisers,” and 65% for snacks, across 30 tests in the first half of 2025. Heritage Kellanova reported an incremental return of 2.4 to 2.9 times its spend across three studies. At launch, the tool was open only to select advertisers on managed-service Sponsored Products campaigns.
In October 2025 DoorDash had made ghost ads its standard for measuring restaurant Sponsored Listings. In “eight recent large-scale tests,” it said, the method reduced measurement noise by over 90% and showed higher incremental sales from Sponsored Listings than earlier tests had found.
The ad business behind those tests has grown quickly. DoorDash and Wolt passed an annualized ad revenue run rate of over $1 billion in 2024, and in June 2026 DoorDash said its ad platforms across DoorDash, Wolt and Deliveroo support more than 400,000 advertisers. It doesn’t report ad revenue separately in its filings.
What pay-per-order does and doesn’t guarantee
Restaurants on DoorDash buy Sponsored Listings by the order, not the click. “We only charge them if it leads to an order,” Peter Giordano, general manager of DoorDash Ads’ platform and growth services, told Marketing Dive in October 2025. DoorDash’s help center says the bid is the most you will pay per order, and that you pay “for orders placed within 7 days of clicking your ad.”
So you pay for attributed orders: anyone who clicked and then ordered within a week. Some of them may be regulars who would have ordered anyway, or people who would have found you in the organic results a scroll further down. Pay-per-order protects you from paying for clicks that lead nowhere. It doesn’t protect you from paying for orders the ad didn’t cause. DoorDash’s Sponsored Listings page says businesses earn “$6 for every $1 spent” on ads on average, citing its own 2023 study of 15,700 merchants. The page doesn’t say how much of that the ads caused, and that gap is what ghost ads are built to measure.
How ghost ads measure what an ad caused
To know what an ad caused, you compare people who saw it with similar people who didn’t. Randomly split your audience, show ads to one group, hold the other back, and compare their orders. The difference is the incremental effect. The hard part is building the control group fairly and cheaply, and the research compares three main designs.
In a public service announcement test, the control group sees a charity ad in your slot. That works, but you pay for those impressions, and the ad system may deliver the charity ad to different people than your real ad. In an intent-to-treat test, you compare everyone assigned to each group, whether or not they actually saw the ad. That adds no ad cost, but many people in the ad group never see the ad, often because it sits too far down the page, so the effect gets diluted and the result is noisy. DoorDash says intent-to-treat “often underestimates the real ROI” of campaigns.
Ghost ads solve both problems. When a control-group user would have been shown your ad, the system records a “ghost” impression and shows them something else, organic content on DoorDash or the next ad in the auction elsewhere. That lets you compare people who saw your ad with people who would have seen it, and leave out everyone who was never in the running. DoorDash says real advertisers “never bid against ghosts,” and that the method cut experiment dilution by 92% and tightened confidence intervals on incremental return on ad spend by 35%.
DoorDash adopted the method from published research. Garrett Johnson, Randall Lewis and Elmar Nubbemeyer described it in the Journal of Marketing Research in 2017, using Google’s display network. In their retailer test, ads lifted website visits by 17.2% and purchases by 10.5%, and they found advertisers could measure lift just as precisely while spending at least an order of magnitude less than intent-to-treat or charity-ad tests required. Google’s Display & Video 360 uses the same idea in its Conversion Lift studies, where control users “are shown the next ad in the auction.”
Why attributed results mislead in both directions
In the best-known studies, attribution has overstated what ads did. When eBay paused its brand-keyword search ads on Yahoo! and Microsoft’s search engine, natural search results recovered “almost all (99.5 percent)” of the lost clicks, and the researchers found the ads had no measurable short-term benefit. A study of 15 US advertising experiments at Facebook compared randomized tests with observational methods common in the industry and found the observational methods often failed to produce the same effects. In one study, the experiment measured a 73% lift while a simple exposed-versus-unexposed comparison suggested 316%.
A badly designed test can understate effects too. DoorDash’s ghost-ad tests found more lift from Sponsored Listings than it had measured before, and its technical write-up says intent-to-treat designs often underestimate returns. A clean holdout can move your number in either direction. What it gives you is an estimate of what the ads caused, with a margin of error you can see, and that is the figure to budget against. Our guide to measuring beyond last click applies the same thinking to Meta.
What to ask any ad platform about incrementality
When platforms publish lift figures, such as DoorDash’s 30% or Instacart’s “more than 15%,” the tests are their own. Before you rely on one, ask how the control group is built: ghost ads, charity ads or intent-to-treat. Ask what share of your audience is held back, how long the test runs, and which orders count. Ask for incremental return on ad spend instead of attributed return, and multiply it by your profit margin: if the result is below 1, each ad dollar loses money. And ask whether you can run the test on your own account. At its March 2026 launch, DoorDash’s Sales Lift Measurement was open only to select advertisers on managed-service campaigns.
The same questions apply to Amazon, Walmart and every other network covered in our guide to retail media, and to the clean rooms retailers use to match your data with theirs, which our look at AWS and Amazon advertising explains. For a single restaurant, the closest version is a pause test. Switch Sponsored Listings off for a few weeks, or in some of your locations, and compare total orders with the weeks or locations where ads ran, keeping in mind that seasons and promotions move orders too. Local business marketing covers the local search and listings side of a restaurant’s visibility, and pause or holdout tests like this one belong in any performance marketing plan.
Frequently Asked Questions
What are DoorDash ghost ads?
Ghost Ads is the method DoorDash uses to measure whether its ads cause extra orders. Customers are randomly split, and when someone in the control group would have seen an ad, the system logs a “ghost” impression and shows organic content instead. Comparing people who saw the ad with people who would have seen it isolates the ad’s effect. DoorDash made it standard for Sponsored Listings in October 2025 and in March 2026 began offering it to consumer brands through Sales Lift Measurement.
How does DoorDash charge for ads?
Sponsored Listings, the restaurant ad product, are priced per order rather than per click. Advertisers set a bid, which is the most they will pay per order, and pay for orders placed within seven days of a customer clicking the ad. That means advertisers pay for attributed orders, which can include customers who would have ordered anyway. Sponsored Products, the ads consumer brands buy, are priced per click instead.
What is incrementality testing in advertising?
Incrementality testing measures the sales or orders an ad caused, as opposed to the ones it was credited with. A random group of people is held back from seeing the ads, and their purchases are compared with those of people who saw them. The difference is the incremental effect. Common designs are ghost ads, public service announcement controls, intent-to-treat tests and geographic tests.
How big is DoorDash’s ad business?
DoorDash said its ads, together with Wolt, passed an annualized revenue run rate of over $1 billion in 2024, and in June 2025 it said they served more than 150,000 advertisers. In June 2026 it said its ad platforms across DoorDash, Wolt and Deliveroo support more than 400,000 advertisers. DoorDash doesn’t break out ad revenue in its financial filings, though executives have said ads are a growing share of the business.



