Roku Curate and Retail Partnerships Drive Shoppable CTV Results

Full analysis

Roku wants to sell you a closed loop: 100 million logged-in households on one end, a retailer's purchase data on the other, and Roku's inventory in the middle. Carly Friedman pitched Roku Curate at Beet.TV with a March Madness campaign for The Hershey Company's Reese's, run with Instacart, that reportedly moved 30% more new buyers and returned more than $4 for every dollar spent. The question for operators is not whether that campaign worked. It's whether the machinery under it repeats when the brand isn't a universally-known candy and the moment isn't the biggest sports window of the year.

What's being decided: whether CTV retail media off the walled gardens is a real budget line for publishers and buyers, or a set of logo-partnerships that stay small. Easy to undo for a buyer testing one campaign. Hard to undo for a publisher betting its data strategy on it. Nothing sets a hard deadline except the Q4 upfront commitments already being written.

The Market Analyst. This isn't a new category. It's Roku building toward where Amazon Ads has lived for three years: owned inventory, a retail purchase signal, closed-loop attribution in one place. In plain terms, Roku is trying to prove that money now spent on Amazon and Meta for performance can move to the biggest screen in the house. The signal that decides it is the partner list. If Walmart Connect or Target Roundel shows up on Curate, that's a genuine off-site threat to Amazon and worth repricing Roku's ad story around. If it stays Best Buy and Instacart, this is electronics and grocery, a niche, not a platform. AppLovin set the bar in mobile: repeatable closed-loop returns at scale is what pulls budget. One case study doesn't clear it.

The Skeptic. A 30% new-buyer lift for Reese's during March Madness is the easiest lift you can buy. Take a snack brand everyone already knows, add the highest-attention sports window of the year, add Instacart's purchase data, and lift is close to guaranteed. Strip out any one of those and I want to see the number again. The harder question nobody answered: what's the match rate between Roku's 100 million logged-in households and a given retailer's loyalty file? If it's 40%, the real addressable base is far smaller than the headline. And Curate needs retailers to hand over purchase data at a granularity most legal and procurement teams fight. Best Buy and Instacart are logos. They are not proof of scale.

The Operator. For a media buyer this is a workflow win before it's a data win. Curate pre-wires measurement with Best Buy and Instacart, which means fewer insertion orders, fewer discrepancy calls, fewer reasons a retail campaign stalls in setup. That's real. The thing that breaks around day 90: reconciliation. When a brand's own marketing-mix model, the top-down math that credits channels for sales, doesn't match Roku's closed-loop number, somebody on the activation team owns that meeting. Anyone committing Curate dollars into Q4 needs a reconciliation protocol agreed with the client before the first invoice, not after. Otherwise the $4 return becomes a fight, and the fight kills the renewal.

The Customer / End User. Two customers here, and they want different things. The brand, Hershey's side, wants proof that holds up when the CMO's mix model is watching, not a sports-week highlight. The retailer, Best Buy and Instacart, is being asked to share its most valuable asset, first-party purchase data, in exchange for ad demand it doesn't fully control. That's the tension nobody at the podium mentions. Retailers built media networks to keep that data and the margin. Handing it to Roku's closed loop only makes sense if Roku brings demand the retailer can't reach alone. For mid-tier retailers with no DSP of their own, that trade is attractive. For Walmart, which owns its inventory and its shopper, it's a reason to stay away.

The CFO. The $4 return is a pilot number under ideal conditions. I don't underwrite a budget on it. The cost that doesn't show in the deck is the data plumbing: match-rate leakage, the legal cycle to get a retailer comfortable, the analyst time spent reconciling two attribution systems that disagree. At scale, closed-loop CTV pays back only if the match rate holds above the level where the audience gets too thin to matter, and only if the returns repeat across brands that aren't Reese's. Payback isn't the campaign. It's the second and third campaign coming in at the same efficiency without a tentpole carrying the result.

Where they part ways. The Market Analyst says the partner list is everything. The Customer says the partner list is exactly what won't fill out, because the retailers with real scale have every reason to keep their data and inventory to themselves. That's the live disagreement: is Curate an on-ramp for retailers who can't build their own stack, or a product structurally capped because the retailers worth having don't need it? Second split: the Operator sees a genuine setup improvement worth adopting now; the Skeptic says the setup is easy and the proof is not, and buyers are being sold the easy part.

What it hinges on. Two facts, neither disclosed. The match rate between Roku's logged-in households and a given retailer's loyalty data, which sets the real reach. And whether the returns repeat off a tentpole with a brand that isn't already universally known. Everything else is packaging. Before committing Q4 dollars, buyers should demand a second case study, non-sports, non-CPG-staple, and a stated match rate for their specific retailer. Publishers weighing their own retail-data strategy should watch who signs, not what lifted.

The council leans skeptical on the scale story and positive on the workflow. Curate is a real convenience and a thin proof.

Prediction: Roku will not name any additional retail partner for Curate — the product that matches Roku's logged-in TV households to a retailer's purchase records — beyond Best Buy and Instacart before the 2027 upfront season closes on May 31, 2027.

Confidence: Medium

Why: Curate's public proof so far is one campaign: Reese's during March Madness with Instacart, reported at 30% more new buyers and more than $4 back per dollar spent, which is the easiest possible test case — a famous snack brand in the biggest sports window of the year. Retailers built their own media businesses precisely to keep purchase data and the margin that comes with it, so the only ones likely to hand that data to Roku are those with no ad sales machinery of their own, a shrinking pool once Best Buy and Instacart are taken. Signing one costs a retailer months of legal and procurement work and an unproven match rate between Roku's roughly 100 million logged-in households and its own loyalty file, against demand Roku has not yet shown it can deliver away from a tentpole event. The opposite outcome — a steady drip of new retail logos — is what a product launch normally produces, which is why calling zero additions is the real test of whether Curate is a platform or a pair of partnerships.

Revisit by 2027-05-31: We're right if Right if, as of May 31, 2027, Roku's public announcements and product materials list no retail data partner for Curate other than Best Buy and Instacart. We're wrong if Wrong if Roku publicly names any additional retail data partner for Curate on or before May 31, 2027.