The Loyalty Spillover: How a Food Delivery Subscription Feeds the Restaurant Down the Block

Based on the research ofJaeho Myeong, Yongjin Park, Francis Joseph Costello, Keongtae Kim, and Jae-Hyeon Ahn, "From Bytes to Bites: Examining the Online and Offline Spillover Impacts of Online Food Delivery Platform Subscriptions," Management Science, 2026

The Subscription Was Supposed to Just Steal Share

The standard story about platform subscriptions is a zero-sum one. A customer pays a flat monthly fee, the waived delivery charge nudges every order toward the subscribed app, and the money that used to flow to a competing delivery platform now flows to you instead. Jaeho Myeong, Yongjin Park, Francis Joseph Costello, Keongtae Kim, and Jae-Hyeon Ahn confirm that first part using a large-scale individual-level transaction data set and a difference-in-differences design around the launch of an online food delivery subscription service. Subscribers did shift their online food delivery spending away from a competing platform toward the focal one.

What the standard story gets wrong is the ending. The authors find no cannibalization of the overall online food delivery market: total consumption within the category went up, not just moved around inside it. And the growth did not stop at the edge of the app. Using geospatial analysis of where subscribers actually went, the paper shows subscribers became more likely to dine out locally and spend in nearby, noncompeting sectors of the local economy, activity the platform never processes a single transaction for. A loyalty mechanism built to keep someone inside one app ends up pushing real foot traffic into restaurants that app cannot even see.

Why Lock-In Produces Wandering, Not Just Repetition

The mechanism the authors isolate is not simple habit. It is variety-seeking, constrained by geography. Subscribers were more likely to try offline restaurants that predominantly serve cuisines the OFD market itself does not carry well, and that pattern held mainly within a proximal radius of the subscriber's home or usual delivery area. Push the same restaurant further away and the effect fades, which the paper attributes to transportation costs. A subscriber freed from delivery fees on their online orders does not become a more loyal customer of any single restaurant. They become a cheaper experimenter, and the cheapest place to experiment outside the app is whatever is already a short walk or drive away.

This is a useful correction to how most platform operators think about retention metrics. A rising repeat-purchase rate inside the app looks like loyalty. The paper suggests it can just as easily be the leading indicator of a subscriber with more disposable attention and a lower marginal cost of trying something new, some of which they will spend on you and some of which they will spend three doors down from the restaurant you deliver for.

A subscriber freed from delivery fees does not become more loyal to any single restaurant. They become a cheaper experimenter, and the cheapest place to experiment is whatever is already a short walk away.

DashPass and Uber One: The Same Shape, in the Real World

Nobody needs to take this mechanism on faith. The major US delivery subscriptions already show pieces of the same pattern, even though they were not the data set behind this particular paper. DoorDash's own reporting on DashPass usage found that 83 percent of surveyed DashPass members said they had used the app to try a restaurant they otherwise would not have, compared with 65 percent of non-subscribers, and that members were also more likely than non-subscribers to try a new retail or grocery merchant inside the app, 23 percent versus 12 percent. That is self-reported, promotional data from the company that runs the program, not an independent audit, but it is directionally the same claim the Management Science paper makes with transaction records: a subscription fee that removes delivery friction increases exploration, not just repeat ordering of the same thing.

The scale of lock-in these programs have built is large enough that any spillover, even a modest one, compounds into real local activity. DoorDash disclosed more than 18 million members across DashPass and Wolt+ by the end of 2023, and Uber has reported roughly 19 million Uber One members across 25 countries. Uber's delivery platform passed 50 million monthly active users for the first time in September 2024, by the company's own account. On a 2022 earnings call, Uber's then-CEO Dara Khosrowshahi told analysts that Uber One members accounted for about 23 percent of the company's overall gross bookings and roughly 32 percent of delivery bookings specifically, a sign of how concentrated spending becomes once someone is inside the subscription. The Management Science paper's contribution is showing where some of that concentrated spending habit actually leaks: not just to a competing app, but out the front door entirely.

The local-economy numbers these companies report in parallel, while not a measure of the subscription mechanism itself, show how large the surrounding base already is for spillover to work against: DoorDash's own 2024 US economic impact research, commissioned from outside research firms, puts merchant sales on the platform at more than 40 billion dollars for the year across over 550,000 local merchants. That is the backdrop of local commercial density the Management Science paper's geospatial result depends on. Spillover needs somewhere nearby to spill into.

The Transportation Ceiling on Every Loyalty Program

The paper's most useful constraint for operators is the one that limits the upside: the spillover is geographically bounded. It shows up in nearby restaurants and nearby sectors, within a radius the subscriber will actually travel without a car trip that erases the convenience the subscription was sold on. That ceiling matters because it tells you the spillover is not a general halo effect of your brand. It is a local-density effect, present only where there is enough variety within walking or short-driving distance to seek out in the first place. A subscriber in a dense urban corridor with dozens of restaurants nearby has somewhere to wander. A subscriber in a low-density suburb largely does not, no matter how attached they are to the app.

That distinction matters for anyone trying to extend this finding to other subscription categories. Amazon Prime's two-day shipping plausibly increases how often a member runs errands near their home while waiting for other things, Netflix's recommendation engine plausibly increases variety within streaming without ever sending anyone outside, and transit passes plausibly increase incidental local spending along a fixed route. Each of those is a hypothesis, not a finding, since none of them were tested with this paper's geospatial method. What the Management Science study actually establishes is narrower and more useful: in food delivery specifically, a subscription's offline spillover rides on local restaurant density and travel cost, and a platform operator who wants to claim credit for growing the local economy around their service needs the density to make that claim true, not just the subscription.