If you're launching or scaling a retail media network, you're going to hear this advice early:
"You need a CDP."
Sometimes it's said by CDP vendors. Sometimes by agencies. Sometimes by well-meaning consultants who've seen CDPs work well in brand marketing stacks.
But for retail media, a CDP is often optional—and in many cases, it's a detour that slows down the thing that actually drives RMN success: monetization performance + operational scalability.
Retail media isn't an email program. It's not a personalization project. It's not a "profile warehouse."
Retail media is an economic system: supply, demand, auctions, yield, pacing, measurement.
And good RMNs, including many powered by Topsort, do not treat a CDP as a prerequisite at all.
Why retail media keeps getting pushed a CDP
Here are the common reasons people recommend a CDP, and why those reasons often don't justify going “CDP-first.”
"You need audience targeting."
In retail media, the most reliable targeting inputs at launch are usually search intent, category and product context, onsite behavioral signals (e.g., views, add-to-cart, purchases), seller and brand constraints, and simple retargeting windows, where privacy allows.
But those things don’t require a full CDP. They require clean event collection and a segmentation layer that can activate into the ad system.
"You need identity resolution."
Most RMNs don't need a grand unified profile to start monetizing. They need a consistent, privacy-safe way to represent users and sessions, basic consent controls, and stable measurement definitions. Identity resolution becomes more important when you're doing sophisticated offsite activation across channels, and that's not a day-one problem.
"You need measurement."
Retail media measurement isn't a question of "did the email get opened." It's what happened after the impression or click, incremental lift when you're ready for it, and attributed outcomes tied to retailer-native signals. That depends far more on well-instrumented commerce events and clean data joins than on CDP profile stitching.
The core mistake: confusing marketing stacks with RMN stacks
CDPs were built for a specific world: unify customer profiles, power owned and earned marketing, personalize experiences, orchestrate lifecycle messaging. That's a legitimate job. It's just not the job retail media needs done.
Retail media's job is different: run auctions reliably, optimize yield over time, price inventory correctly, create a self-serve system advertisers can scale, and produce trusted reporting tied to real outcomes.
A CDP can be useful later, but it's not the core monetization engine. Treating it as one is where teams go wrong.
What you actually need to succeed in retail media
A scalable RMN stack looks more like this:
A strong auction and monetization engine. This is the foundation—retail-native auction logic, pacing and budget controls, floor strategy and yield optimization, and learning loops that improve performance over time. This is where revenue is created and compounded.
Clean commerce signal collection. You need product and catalog context, inventory awareness where possible, conversion events, and consistent definitions that power reporting. The goal isn't profiles. It’s signals that improve auctions and measurement.
A lightweight audience and segmentation layer. You can build effective audiences without a CDP by defining segments directly from first-party events, applying privacy-safe rules, and activating those segments into onsite placements first. Offsite expansion can come later, when the business warrants it.
A single source of truth for reporting and billing. RMNs break when the serving data says one thing, analytics says another, and finance says something else. This isn't a CDP problem — it's an architecture problem, and it needs to be solved at the infrastructure level.
What the "CDP-first" trap actually costs you
When RMNs put a CDP at the center of the architecture too early, the common outcomes are slower time-to-launch, more vendors to coordinate and more failure points, batch latency problems in systems that need real-time decisions, operational overhead as ad ops teams fill the gaps, and fractured optimization loops that never fully compound.
The irony here is this: many teams buy a CDP to “get more sophisticated,” and end up less scalable because the monetization engine was never the focus.
When a CDP does make sense for retail media
A CDP can be genuinely valuable when you have clear, proven needs. Things like: advanced offsite audience activation across channels, complex lifecycle personalization requirements outside the ad system, strict cross-property identity goals across multiple brands or sites, or heavy CRM and marketing orchestration that lives beyond the RMN.
But even then, the healthiest pattern is usually monetization infrastructure first, CDP later if needed—not the other way around.
A simple decision framework
When you inevitably hear that CDP advice, ask one question:
Will a CDP increase RMN revenue this quarter, or just increase complexity?
If you're still proving product-market fit for ads, focus on launching fast, getting auction and measurement right, compounding performance, and building advertiser trust. That's how strong RMNs are built.
And it's also why many high-performing retail media businesses, including Topsort-powered ones, don't see a CDP as required at all.
The bottom line
Retail media is transitioning from tools to infrastructure.
A CDP is a tool, useful in the right place, but it is not the infrastructure layer that makes retail media scale.
If you want a durable RMN, build a system that optimizes monetization, reduces manual operations, and compounds performance over time.
That's Retail Media 3.0.