The On-Site Habit:
Retail Media Has Moved, Most Budgets Haven't

Your Retail Media Budget Is Probably Backwards
Most retail media budgets were built when on-site was the only real option. They haven't been meaningfully revisited since, even as the place where retailer data actually gets applied has shifted pretty dramatically.
The brands getting the best incremental returns right now aren't necessarily spending more. They're spending differently, and the difference usually comes down to how much of the budget is still parked on-site out of habit versus how much is following the actual inventory shift.
What the numbers are actually saying
Off-site retail media is growing roughly three times faster than on-site formats. That gap isn't a prediction, it's current spend behavior from brands that have already run the comparison. The reason is straightforward: on-site inventory in most major categories is saturated. Sponsored product CPCs have been climbing for two years. The incremental return on additional on-site spend keeps compressing because you're reaching the same in-market shoppers your competitors are also bidding on, at prices that reflect how crowded that auction has gotten.
Off-site is a different supply situation entirely. Walmart's Vizio acquisition, Amazon DSP's access to Netflix, Tubi, and Roku, and grocery networks expanding into streaming have created real scale in CTV inventory that didn't exist three years ago. The targeting is the same retailer shopper data. The measurement, closed-loop back to actual transactions, is now standard at the major networks. The inventory is just less crowded and the audiences have lower overlap with what you're already reaching on-site.
The budget conversation most teams are avoiding
The default in most planning cycles is to treat off-site retail media as a test budget, something that gets a small allocation after the on-site plan is locked. Off-site was a reasonable test budget two years ago when measurement was thin. The closed-loop infrastructure at the major networks has since caught up, and brands that ran the incremental ROAS comparison moved budget accordingly.
The brands that have moved meaningful budget off-site this year are doing it because the incremental ROAS comparison came back in favor of off-site, and holding the line on on-site allocation at that point is a deliberate choice to overpay for diminishing returns. Before next quarter's plan is locked, the more useful question is how much of your current on-site spend is actually incremental versus how much is defending visibility you're already paying for at a premium price.
What actually has to change
Most retail media reporting keeps on-site and off-site as separate lines. The interaction between them is where most of the lift shows up, and brands optimizing each in isolation are making allocation decisions on partial information.
Creative is the other piece. Off-site retail media is video-first and needs to be built for the format. Repurposing sponsored brand assets doesn't work. The brands seeing strong off-site results are producing CTV-native creative tied to the same shopper signals their on-site campaigns are already using, and with production costs down significantly across the major networks, the overhead argument for staying on-site is weaker than it used to be.
The allocation comparison we’re seeing across clients this year is fairly straightforward: incremental ROAS from additional on-site spend versus incremental ROAS from off-site. For most brands running this analysis, off-site is outperforming, and budgets are shifting in that direction accordingly.
The on-site foundation still matters. It just shouldn't be eating a budget share that made sense three years ago.