The Infrastructure War: Owners, Builders. Renters.
Part 2: Three tiers of retail/commerce (RMN/FMN) media infrastructure. One is closed. One is compounding right now. One is paying a hidden tax on every campaign. Which one are you?
Part 1 mapped the landscape - RMN, CMN, FMN, the $200B shift, what AI is actually doing. If you haven't read it, start there. This one picks up where that left off.
Most teams still see inventory. The smarter ones see infrastructure. The winners will see decision systems.
That last line from Part 1 wasn’t a prediction. It’s a diagnostic. And when I teased this piece as “Renters vs. Owners,” I was giving you the simplified version. The full picture is more useful. And more uncomfortable.
It’s not a binary. It’s a spectrum. Three tiers, one direction of travel and the tier you’re in determines whether your infrastructure compounds or quietly decays under you.
Owners. Builders. Renters.
The Permission Layer: Consumer Love Is the Entry Fee, Not the Moat
Every operator who’s built one of these networks lands on the same truth: none of this works without consumer trust. Amazon’s $60B in ad revenue isn’t about targeting technology it’s because 200 million Prime members are in an active, reciprocal relationship with the platform. Costco members tolerate a credit card pitch at checkout because that trust has been earned over decades, not engineered in a campaign brief.
This is the permission layer. Without it, you’re selling overpriced impressions with a first-party label. Consumer love is table stakes. What you build on top of it is the actual war.
Owners. Builders. Renters. Where Does Your Stack Actually Sit?
When Part 1 mentioned 277 retail media networks worldwide, the natural reaction is: how do 277 networks compete? Most of them don’t. Behind every retailer’s network is an ad tech stack. Some built their own. Most assemble from vendors: The Trade Desk, LiveRamp, Epsilon, Criteo, Quotient. Different logos. Same pipes. Three tiers and the tier you’re in determines everything.
Tier 1 — The Owners. Amazon and Walmart. This tier is closed. A decade of building proprietary DSPs (the technology that buys and places ads programmatically), identity graphs, AI, and CTV surfaces- Walmart Connect reported 53% year-over-year global ad growth in Q3 2025, well ahead of Amazon’s 24%. The window for in-house-everything closed around 2018. Owners aren’t a model to aspire to. They’re a ceiling to understand.
Tier 2 — The Builders. This is where the real war is being fought and it’s the tier still open. Builders assemble a composable stack: a governed data warehouse, modular identity resolution beyond email-matching, always on audience enrichment, and bidirectional DSP connectors that close the loop between data and activation (more on this below). Target’s Roundel, Kroger’s Precision Marketing, McDonald’s, Nike - Builders don’t own everything. They control what matters.
Tier 3 — The Renters. Everyone else. No standard attribution model, no standard conversion definition, no standard reporting format. Comparing ROAS (return on ad spend) across four networks is still a manual exercise. Most Tier 3 companies haven’t made the infrastructure decision yet and they’re paying the compounding cost every quarter with no line item to show for it.
The Owners tier is closed. The real war is between Builders and Renters. The prescription isn’t: become Amazon. It’s: stop renting. Start building.
The Renter’s Tax: What Borrowed Infrastructure Actually Costs
Most Tier 3 brands pay in three ways that never appear on an invoice.
Match rate decay. As third-party signal degrades, audience match rates drop quietly and consistently. Campaigns just work less.
CPM inflation. As signal quality drops, bid density required to hit reach targets goes up. You spend more to reach the same person.
Attribution drift. The measurement layer justifying your spend is measuring its own noise and calling it performance. 36% of marketers cite difficulty proving incrementality as the primary reason they’d reduce retail media investment.
This is the identity tax. You won’t get an invoice. You’ll notice efficiency declining — and your agency will recommend more spend to compensate. That recommendation is the tax being collected.
What the Builder’s Stack Actually Looks Like
The foundation is a governed data cloud treated as an activation engine, not a reporting layer.
The differentiator is identity resolution. Email-based matching works only on the audience you already know. Builders construct identity graphs beyond their first-party footprint: probabilistic matching (connecting customer signals even without a direct login or email), household graphs (linking individuals to the actual households making purchase decisions), device graphs (stitching together the phone, laptop, and tablet that belong to the same person) woven into a persistent identity that survives cookie loss and channel fragmentation. They’re reaching audiences their competitors literally cannot see.
Then activation connectivity - bidirectional DSP connectors that push enriched audiences out to media platforms and pull performance signals back in so every campaign informs the next one. Every dollar spent sharpens the next decision.
Then decision traces structured records linking signal to decision to action to outcome. Institutional intelligence that doesn’t walk out the door when a campaign manager leaves or a contract ends. Only 12% of commerce media decision-makers have reached full-funnel capability across on-site, off-site, and in-store. They built earlier. The gap compounds in their favor every quarter.
The Managed Middle — and the Blurry Lines
Within the Renter and Builder tiers, many companies use managed solutions — third parties running their commerce media infrastructure end-to-end. Worth naming separately because it’s the most common path, and the most misunderstood.
It’s rational for many. Thrive Market chose Instacart’s Carrot Ads over building their own: “For our size and scale, it doesn’t make sense to reinvent the wheel.” Fair. But managed solutions raise one uncomfortable question: does it compound — and for whom?
When Amazon syndicates its ad infrastructure to Macy’s, Amazon gets richer signal and a compounding identity graph. Macy’s gets revenue. The compounding happens on one side. Retail media and membership now account for roughly one-third of Walmart’s operating income. If advertising margin is a core P&L line, outsourcing the infrastructure generating it is operationally equivalent to outsourcing your margins.
Starbucks owns its AI personalization layer. McDonald’s runs loyalty infrastructure closer to Spotify than a QSR. Nike pulled out of Amazon to own the identity relationship directly. Builder decisions — all of them. The brands defaulting to traditional operator logic are staying in the managed middle, renting from companies compounding faster than they are.
Five Questions Every CMO, CDO, and GM Should Be Asking And Every Practitioner Should Be Able to Answer
For executives: due diligence before the next planning cycle, RMN pitch, or DSP renewal. For practitioners - Retail Media Managers, Measurement Leads, First-Party Data Strategists these are your ammunition. Knowing your organization’s tier is career context, not just strategic context.
1. What is our actual match rate and how has it trended over the last four quarters? If no one knows, you’re paying the identity tax without knowing the amount.
2. Is our first-party data being activated or just stored? A governed data cloud with no activation layer is an expensive hard drive.
3. How many hops exist between our customer data and our media execution? Every hop is a tax on match rate, latency, attribution fidelity, and data governance risk.
4. Are we measuring what we’re influencing or only what we can attribute? If you can’t run a holdout test, you don’t have measurement. You have correlation labeled as performance.
5. If our DSP contract ended tomorrow, what do we own? Audience lists? Identity graphs? Decision traces? Or just a login screen and a rate card?
The answers tell you your tier. Most teams don’t like what they find. But asking is already halfway to deciding.
What This Means for Your Career
The tier your organization sits in compounds for them and for you. A Retail Media Manager at a Tier 3 Renter optimizes on borrowed infrastructure with no learning loop. A Retail Media Manager at a Tier 2 Builder operates inside a compounding system — decisions get smarter, measurement gets cleaner, work is defensible in ways vendor-dependent campaign management never is.
Skills compound too. Activating against a governed identity graph, running real holdout tests, closing the loop between warehouse and DSP — these Builder-tier skills don’t develop inside a Renter stack. And they’re exactly what the market is paying a premium for. The question “which company should I join?” now has an infrastructure answer. The logo on your resume matters less than the stack you learned to operate on.
Most teams reading this are somewhere between Tier 2 and Tier 3. That’s not a failure it’s a starting point.
The gap is infrastructure. More spend doesn’t close it. Better systems do.
The One-Paragraph Summary for Any Meeting
The commerce media infrastructure war isn’t about who has the most networks or the biggest budget. It’s about three tiers: Owners who built proprietary monolithic stacks a decade ago and whose window is closed; Builders assembling composable infrastructure governed data cloud, advanced identity resolution, bidirectional activation that compounds with every campaign; and Renters optimizing on borrowed infrastructure and paying an identity tax they can’t see. The Owners tier is unreachable. The Builder tier is open. The longer a brand waits, the more expensive the transition becomes and the wider the gap gets on match rates, measurement confidence, and decision speed. More spend doesn’t close the gap. Better systems do.
Next week: Part 3 — The Decision Layer. Why the brands winning commerce media aren’t just running better campaigns. They’re making faster, better decisions. Making sure all the 3 stakeholders are happy: 1. End Consumer 2. Brands 3. CFO And what the infrastructure enabling that actually looks like from the data cloud to the activation layer to the learning loop that makes every dollar compound.
Author’s note: An LLM helped with research, citations, and light proofreading without changing content. All ideas, arguments, voice, and em dashes are mine. :)