Dollar General Unifies Onsite and Offsite Retail Media via Trade Desk and Kevel

Full analysis

Dollar General wired its retail media network into a single pipe. The Trade Desk runs the off-property buying, the open web where the retailer's ads chase shoppers around. Kevel runs the ads on Dollar General's own site and app. Now they share one audience file, one set of targeting, one measurement layer. Before, brands ran two campaigns on two platforms and someone stitched the numbers together on the back end. Austin Leonard, who runs Dollar General Media Network, called it the "white whale."

What's actually being decided: not whether Dollar General does this, that's done. The real question for operators is whether "own the on-site ad server, rent the off-site pipes" becomes the default build for every mid-size retail media network. Easy to undo? For Dollar General, no. Migrating to one audience graph and retiring a reconciliation workflow is months of data-team work you don't reverse casually. For the industry, the pattern is still young enough to break. What sets the deadline: the 2027 upfront and annual joint business plans, when CPG brands decide where their retail media dollars go.

The Market Analyst. The quiet winner is The Trade Desk. Every tier-two retail media network that copies this blueprint sends its off-property budget through TTD's platform at almost no new cost to TTD. That compounds. Wall Street is fixated on TTD's connected-TV story and is underrating this retail media on-ramp. The quiet loser is Amazon. Most off-property retail media budgets were defaulting into Amazon's own buying platform; every dollar routed through TTD and Kevel instead is a dollar Amazon didn't touch. Kevel is the asymmetric bet here: a small ad-server vendor that re-rates hard if five more retailers buy the "own your stack" argument. In plain terms: TTD just added a new customer type that costs it almost nothing to serve.

The Skeptic. Calling this a "white whale" is a warning, not a boast. White whales sink ships. Strip the language and what shipped is an API handshake between a Kevel ad server and TTD's platform. Useful, not novel. Here's what didn't change: Dollar General doesn't own shopper identity at the depth Walmart or Amazon does. Its base skews rural, cash, low-digital. And most CPG brands buy Dollar General's network because Walmart Connect is sold out, not because they love the data. Unified measurement only matters if the underlying audience signal is differentiated. If it isn't, they built a nicer dashboard on thin data. A clean number describing a weak audience is still a weak audience.

The Operator. The pain lands at 90 days, and it lands squarely on Dollar General's own data team. The vendors shipped their part. Somebody has to own one data taxonomy and make onsite and offsite agree. That "dirty secret" quote, two attribution models running in parallel, is the whole job. Merge them and CPG brand managers immediately demand unified incrementality reporting, meaning proof the ads actually caused sales rather than counting sales that would have happened anyway. If the schema is dirty, you've traded two broken dashboards for one broken dashboard that more people now trust. And the ad-ops staff who built the manual reconciliation? Their work just got automated. Expect turnover from the people who knew where the old numbers came from.

The CFO. Matt Newcomb of Kevel made the argument that matters for the P&L: retail media revenue is now material enough that renting your entire ad stack to a third party bleeds margin you can't afford. That logic favors owning the on-site server, where the retailer keeps the take. But the off-site side is the opposite call. Building a buying platform to compete with The Trade Desk is a capital sinkhole no tier-two retailer wins. So the real financial thesis is the split: own the piece with durable margin, rent the piece that's commoditized. The math only works if Dollar General's scale, roughly 38,000 stores, is big enough to keep brands buying without a custom, expensive integration for each one.

Where the council splits. The Market Analyst sees a compounding land-grab for TTD and Kevel. The Skeptic sees a routine integration dressed in whale language over a shallow data pond. Both can be right: the architecture can spread across the industry while Dollar General specifically stays a fill-in buy for brands locked out of Walmart. The second disagreement is on margin. The CFO and Strategist buy Newcomb's "own your stack" logic; the Skeptic counters that owning a stack on top of weak identity just protects the wrong thing well.

What it hinges on. One belief: does the "own on-site, rent off-site" split actually hold as the standard build for mid-size retail media, or does it stay a Dollar General one-off? The unified measurement is real plumbing. Whether it drives incremental CPG dollars depends entirely on audience signal Dollar General can't manufacture from a press release. Before treating this as a template, verify that a second, unrelated retailer ships the same architecture with a different DSP or the same one. One deal is an anecdote. Two is a pattern.

Prediction: By the close of Q2 2027 earnings (August 2027), at least two additional US retail media networks beyond Dollar General will publicly announce unified onsite-plus-offsite buying built on an owned ad server paired with The Trade Desk.

Confidence: Medium — the architecture is cheap to copy and the P&L logic is real, but timing depends on retailer roadmaps.

Confidence: Medium. The architecture is cheap to copy and the P&L logic is real, but timing depends on retailer roadmaps.

Why: The signal in this story is that the split-stack design, own the on-site server through a vendor like Kevel and route off-site through The Trade Desk, is now live and being marketed as a repeatable pattern rather than a bespoke project. The mechanism is straightforward: building an in-house buying platform to rival The Trade Desk is a capital sink no tier-two retailer wins, while an owned on-site server protects the margin retail media contributes to the P&L, so every retailer running fragmented onsite and offsite tools faces the same board question Dollar General just answered. The opposite outcome, no retailer follows, would require mid-market networks to keep eating reconciliation overhead they've already admitted is a "dirty secret," which cuts against their own cost pressure.

Revisit by 2027-08-31: We're right if two or more US retail media networks announce unified onsite/offsite buying pairing an owned ad server with The Trade Desk. We're wrong if fewer than two do, or if the copycat deals route off-site through Amazon's platform or an in-house DSP instead.