The convergence of retail media networks and connected TV platforms is accelerating faster than most brands’ measurement strategies can keep pace with. Roku’s launch of Roku Curate — integrating purchase data from Best Buy, Instacart, and other major retailers directly into its CTV ad-buying workflow — follows identical plays by Amazon (through Prime Video Ads and Amazon Marketing Cloud), Walmart (via Walmart Connect’s partnership with Disney and Roku itself), and Kroger (through its 84.51° data science arm feeding into programmatic CTV). The result is a rapidly consolidating landscape where closed-loop measurement — the ability to connect an ad impression to a verified purchase — is increasingly controlled by a small number of platform-retail alliances. For brands, this consolidation creates a real and growing risk: rising CPMs inside walled gardens, diminishing negotiating leverage on measurement transparency, and dependency on a single platform’s attribution methodology to justify spend. This whitepaper lays out the structural dynamics at play, quantifies the risk, and argues that the most defensible response is building a portable, multi-graph measurement framework — one that includes offline channels like programmatic direct mail — before platform lock-in makes that option significantly more expensive.
The Challenge: Closed-Loop Measurement Is Becoming a Platform-Controlled Asset
The promise of retail media has always been straightforward: target ads using purchase data, then close the loop by measuring whether those ads drove incremental purchases. That value proposition is now migrating wholesale into CTV. When Roku integrates Best Buy’s and Instacart’s transaction data into its ad platform, it’s not just offering better targeting — it’s offering a self-contained measurement loop where the platform that serves the impression also owns the conversion signal. This is the same structural move Amazon pioneered with Sponsored TV and Amazon Marketing Cloud, and it creates an asymmetry that brands need to understand clearly.
Here’s the core problem: when the platform controls both sides of the measurement equation — impressions and conversions — the advertiser loses the ability to independently validate performance. eMarketer projected retail media CTV spend would reach $8.2 billion by the end of 2025, and early data from 2026 suggests that number is tracking ahead of forecast. As more ad dollars flow into these closed ecosystems, CPMs inside retail-CTV alliances have risen accordingly. GroupM’s Q1 2026 CTV benchmark report showed that retail-data-enriched CTV inventory commands a 35–50% CPM premium over standard programmatic CTV. Brands are absorbing that premium with limited recourse.
The deeper structural issue is data portability — or the lack of it. A brand running campaigns through Roku Curate can see Instacart conversion data within Roku’s reporting interface, but it cannot extract that conversion data and normalize it against performance from Amazon’s ecosystem, Walmart Connect, or any offline channel. Each platform’s measurement is self-referential. For brands running multi-channel acquisition strategies, this means they’re comparing apples to oranges across every platform report, with no common denominator to determine true incremental lift.
The Solution: A Portable Measurement Framework Anchored in First-Party Data
The antidote to platform-controlled measurement is measurement infrastructure the brand actually owns. This doesn’t mean abandoning retail-CTV platforms — they offer genuine targeting advantages. It means building an independent attribution layer that can normalize outcomes across every channel, including the ones no single CTV platform measures.
The foundation of that layer is the brand’s own first-party data: CRM records, purchase histories, customer addresses, and behavioral signals. When a brand activates its first-party data as the canonical source of truth for who was targeted and who converted, it can run matchback attribution independently of any platform’s reporting. This is where programmatic direct mail plays a structural role that most media plans underestimate.
Direct mail’s attribution model is inherently portable. A matchback analysis compares a known universe of mailed addresses against a known universe of converters — it doesn’t depend on a pixel, a platform cookie, or a retail data clean room to close the loop. When a brand runs programmatic direct mail alongside CTV campaigns in Roku Curate and Amazon Sponsored TV, the direct mail matchback becomes a calibration benchmark: a channel-agnostic conversion signal the brand controls, against which platform-reported conversions can be compared and validated.
This isn’t theoretical. Brands running Postie campaigns alongside digital media consistently use matchback data to audit platform-reported ROAS, identifying cases where CTV or display platforms over-count conversions by 15–30% due to view-through attribution windows that overlap with other channels’ influence. That calibration function becomes exponentially more valuable as more ad spend flows into closed retail-CTV ecosystems where independent verification is difficult by design.
Key Strategies for Brands
1. Treat first-party CRM data as your measurement backbone, not just a targeting input.
Most brands activate their CRM data for targeting — building segments, suppressing existing customers, feeding lookalike models. Fewer treat that same data as the spine of a cross-channel measurement framework. The shift is straightforward but consequential: when every channel’s performance is measured against a single, brand-owned customer file rather than each platform’s self-reported conversions, the brand gains an independent view of incrementality. Programmatic direct mail is natively built on this model — every send is tied to a physical address that maps back to a known customer or prospect record, and every conversion is validated through matchback against that same record. Extending this logic across CTV, display, and paid social means building a unified conversion table where the brand’s CRM is the denominator, not each platform’s pixel.
2. Diversify retail data inputs instead of committing to a single retail graph.
Roku Curate gives you Best Buy and Instacart purchase signals. Amazon gives you Amazon purchase signals. Walmart Connect gives you Walmart purchase signals. None of these graphs overlap neatly, and none of them represent the totality of your customer’s purchase behavior. Brands should resist the temptation to consolidate CTV spend into whichever retail-CTV alliance offers the most convenient data integration. Instead, allocate test budgets across at least two retail-CTV ecosystems and use a channel outside those ecosystems — like programmatic direct mail — as the control. When Postie campaigns targeting the same CRM-derived audience segments consistently produce a 4–8x ROAS with matchback-verified attribution, that becomes the baseline against which platform-reported CTV performance is judged. If Roku Curate reports a 6x ROAS on Instacart-enriched audiences but your direct mail matchback to the same audience shows a 3.5x, the discrepancy isn’t noise — it’s signal that the platform’s attribution window is doing heavy lifting.
3. Build holdout-group discipline into every retail-CTV test.
The gold standard for measuring incrementality — in any channel — is a randomized holdout group. This is operationally simple in direct mail: withhold a random percentage of the mailable audience, then compare conversion rates between the mailed and held-out groups. In CTV, true holdouts are harder because platforms control impression delivery and don’t always support brand-defined holdout segments. This is precisely why multi-channel holdout design matters. A brand can construct a unified holdout: a segment of CRM-matched prospects who receive neither a direct mail piece nor a CTV impression through any retail-CTV platform. Measuring that group’s organic conversion rate against groups exposed to mail-only, CTV-only, and mail-plus-CTV gives the brand a clean incrementality read that no single platform can provide or distort. Postie’s integration with first-party CRM data makes this holdout architecture straightforward — the same audience segments used for mail targeting can be shared as suppression lists with CTV platforms to ensure holdout integrity.
4. Plan for CPM inflation inside retail-CTV walled gardens.
The economics are directional and unfavorable for buyers who don’t plan ahead. As more brands chase retail-data-enriched CTV inventory, the supply of that inventory remains constrained by the number of CTV households and the finite hours of streaming content available for ad insertion. Demand is scaling faster than supply. For brands with CPA targets between $30 and $75 — common in DTC, home goods, and subscription categories — a 35–50% CPM increase can blow acquisition economics within a single quarter. Programmatic direct mail, by contrast, has a fundamentally different cost structure: print and postage costs are stable and predictable, unaffected by auction dynamics or competing bidders. A postcard campaign with an all-in cost of $0.75–$1.10 per piece and a conversion rate of 1.5–3% delivers a CPA that doesn’t fluctuate based on how many other brands are bidding on the same audience. That stability is a strategic hedge against retail-CTV inflation, not just a channel-diversification tactic.
5. Demand data portability in every platform contract.
This is a negotiating posture, not a technology requirement. When signing insertion orders or annual commitments with retail-CTV platforms, brands should explicitly request conversion-level data exports (anonymized and aggregated as necessary) that can be ingested into the brand’s own measurement environment. Some platforms will decline. That refusal is itself useful information — it tells you the platform’s attribution advantage depends on opacity. Brands should weight budget allocation toward partners who support data portability and supplement opaque channels with independently measurable channels like programmatic direct mail, where the brand owns every data point from send to conversion.
Conclusion
The retail media–CTV merger is not inherently bad for advertisers. Targeting ads with verified purchase data and measuring outcomes against that same data is a genuine improvement over probabilistic digital attribution. The problem is structural: when a handful of platform-retail alliances control the entire measurement loop, the advertiser’s ability to independently verify performance, compare across ecosystems, and negotiate on price erodes steadily.
The window to build a defensible, portable measurement framework is open now, but it’s narrowing. As Roku Curate matures and Amazon, Walmart, and Kroger deepen their CTV integrations through 2026 and 2027, the cost of switching measurement approaches or renegotiating data access will rise. Brands that anchor their attribution in first-party CRM data, validate platform-reported outcomes through independently measurable channels like programmatic direct mail, and architect multi-graph holdout strategies will retain the leverage and clarity that single-platform dependence erodes. The brands that wait will find themselves paying more for less transparency — and calling it optimization.
