Eleven Retail Media Networks Is Nine Too Many
eMarketer expects Amazon and Walmart to take 89% of new US retail media dollars in 2026. A four-gate framework for deciding which of your 11 RMNs keep budget and which get cut.
On this page
Start with subtraction. US retail media spend grows from $58.79 billion in 2025 to $69.33 billion in 2026, per eMarketer. That is $10.54 billion of new money. The same firm expects Amazon and Walmart to capture 89% of incremental retail media spending in 2026, and projects that the share of every other RMN it tracks will stay flat or decline through 2027.
Run that out: roughly $9.4 billion of the growth lands at two companies. The remaining ~$1.1 billion gets split across everyone else. Kroger, Target, Instacart, Chewy, Albertsons, DoorDash, Home Depot, and a hundred smaller networks are fighting over about 11% of the new dollars.
Meanwhile, Skai finds the average brand runs six retail media networks today and expects to hit eleven by the end of 2026. So the money is consolidating while the workload is fragmenting. Those two curves are moving in opposite directions, and your P&L sits in the gap.
The overhead nobody puts in the ROAS report
Each network you add is a fixed-cost decision dressed up as a variable-cost one. A new RMN means a new buying interface, a new spec sheet for creative, a new reporting export that doesn't reconcile with the last one, and usually a minimum spend commitment negotiated by someone outside the media team. At EGGKNITE we budget 15–30 hours a month of operator time per active network for a mid-size CPG account. At a blended $85/hour, that is $15k–$31k a year per network before a single impression runs.
Now price the reporting problem. The ANA found 55% of marketers call lack of standardization across platforms their greatest RMN challenge, with attribution to sales close behind at 48%. Eleven networks means eleven definitions of "attributed sale," eleven lookback windows, and eleven dashboards that all claim credit for the same shopper. If you cannot compare networks on equal footing, you cannot manage them as a portfolio. You are just paying eleven separate tolls.
There is also a quieter line item: the political one. An earlier ANA study reported that 88% of brands feel somewhat or heavily influenced by retailers to buy ads on their networks. A chunk of your RMN spend is not media. It is trade relations with a media invoice attached. Fine. Just book it that way.
Four gates before any network keeps its budget
Run every network in your portfolio through these, in order. A network must clear all four or it moves to the cut list.
Gate 1: Incrementality. Can you run a holdout or geo test on this network, or does it at least support third-party incrementality measurement? Platform-reported ROAS on retail media is heavily inflated by demand you already owned; a shopper who searched your brand name on a retailer site was probably buying anyway. If a network only offers its own attributed ROAS and blocks independent testing, its numbers are unauditable. Our primer on what attribution can't answer covers why this gate comes first.
Gate 2: Scale. Can the network absorb 20% more budget next quarter without CPC inflation eating the return? Amazon clears this trivially: eMarketer forecasts it will command $56.71 billion in US retail media ad spending in 2026, with Walmart a distant second at $5.99 billion. Walmart clears it too, and it is the fastest-growing large network; its global ad revenues rose 50% year over year in fiscal Q1 2026, with Walmart Connect up 31%, per eMarketer's coverage of the earnings. A niche RMN where your entire relevant category sees 40,000 monthly searches does not clear it, whatever its ROAS screenshot says.
Gate 3: Data. Does spending here buy you something you act on: closed-loop sales data, clean-room audience access, category share signals you cannot get elsewhere? A network that returns usable first-party signal can justify a lower headline ROAS because the data compounds into better targeting everywhere else. We covered how that signal feeds bidding in the ROAS playbook for AI audience targeting.
Gate 4: Politics. Would cutting this network damage a merchandising relationship worth more than the media waste? Sometimes yes, honestly. If Kroger is 18% of your retail revenue and the buyer expects RMN participation in the joint business plan, that spend is an insurance premium. Keep it, cap it at 5–10% of the retail media budget, and stop pretending it is performance media.
Three budgets, three portfolios
Here is how the gates resolve at three spend levels. Percentages are of annual retail media budget; the test reserve exists so "we might miss something" stops being an argument for keeping nine networks alive.
| Line | $1.5M challenger | $8M mid-market | $40M enterprise |
|---|---|---|---|
| Amazon | 70–75% | 50–55% | 40–45% |
| Walmart Connect | 15–20% | 18–22% | 18–20% |
| Strategic retailer RMNs (politics + data gates) | 0 | 10–15% (1–2 networks) | 20–25% (3–4 networks) |
| Test reserve (one new network per quarter, killed or scaled in 90 days) | 5–10% | 8–10% | 8–10% |
| Total active networks | 2 | 4–5 | 6–7 |
The pattern to notice: even at $40 million, the model tops out at six or seven networks. The average brand is heading to eleven. The difference is nine reporting stacks, five minimum commitments, and roughly $100k+ in annual ops labor buying reach you could have bought better on the two networks where the growth actually is. Pressure-test your own split with the media mix calculator before renewal season locks you in.
The tradeoffs, stated plainly
Pruning is not free. Name the costs before you sign the cut list.
Cutting a captive network can cost you shelf. If a retailer ties RMN spend to merchandising outcomes, walking away has a price outside the media budget. Quantify it with the sales team, then decide. The wrong move is leaving it unquantified and letting fear set the allocation.
Concentrating on Amazon raises your dependence and your auction exposure. When 70% of retail media spend sits with one platform, that platform's CPC inflation is your problem alone. This is the argument for keeping Walmart genuinely funded rather than token-funded, and for revisiting the Amazon versus Google Shopping question annually, since Google remains the main non-retail alternative for product budgets.
A small network you cut may have been quietly incremental. Possible. That is what the test reserve is for: any cut network can re-apply through a proper holdout test. The burden of proof moves to the network, where it belongs.
Consolidation weakens your negotiating story with the survivors. True, slightly. It is outweighed by the leverage of larger commitments on fewer networks, which is exactly the negotiation the big two reward with better rates, betas, and data access.
What to do this quarter
Score all current networks against the four gates. Anything failing two or more gates gets a termination date at the next commitment renewal. Reallocate the freed spend 60/25/15: incremental Amazon budget, incremental Walmart budget, test reserve. Rebuild reporting around one incrementality-adjusted number per network so next quarter's review takes an hour instead of a week; if your current attribution setup can't produce that number, run it through the Attribution Doctor first.
The eMarketer math will not bend for your org chart. 89% of the new money is going to two companies. Your portfolio should already look like you knew that.