Retail media just ate CTV

What a retail media network (RMN) actually does with CTV

A Retail Media Network (RMN) is an advertising platform built on a retailer's own purchase and shopper data. The advertiser buys ad inventory from the retailer, and the measurement uses actual transaction records rather than panel estimates or probabilistic models. Amazon Advertising is the largest global example, with annual ad revenues exceeding $56 billion in 2024 according to its annual report. Walmart Connect, Kroger Precision Marketing, Target Roundel, and Albertsons Media Collective are the US tier-1 equivalents.

The traditional limitation of RMNs was reach. Their inventory sat mostly on-site: sponsored product listings on the retailer's own website and app, display units on category and search results pages. A brand got excellent closed-loop attribution because the retailer controlled both the ad and the purchase, but audience scale was capped at people already browsing within that retailer's property. Independent channel-level verification through marketing mix modelling (MMM) was the standard complement, because on-site RMN attribution could not account for halo effects from other channels.

Extending into Connected TV (CTV) inverts the constraint. Instead of adding more inventory within the retailer's own walls, the retailer takes their shopper identity off-site into streaming environments. A household that shops weekly at Walmart can now be matched to a CTV device and served an ad on a living-room screen, with the purchase outcome tracked back through Walmart's transaction database. The reach gets television scale. The measurement retains retailer-grade determinism.

This is structurally different from how CTV has historically been bought. Standard CTV buys used age-and-gender demographics from the streaming platform, or third-party segment overlays that modelled likely shoppers based on browser behaviour. Neither approach could answer the direct question: did this impression drive a purchase? Retailer first-party data answers that directly, because the retailer owns the purchase record.

The last row matters. Closing the attribution loop is not the same as proving incrementality. A shopper who would have bought regardless of the ad still generates a matched purchase. Both standard CTV and retail media CTV carry this gap unless a proper holdout or geo-split test is layered on top. More on this in section five.

For brands working across e-commerce and retail, this shift is consequential. The measurement language of retail media CTV is the same measurement language the CFO understands: units sold, revenue attributed, return on ad spend (ROAS). The translation from impression to outcome is direct. That changes the internal conversation about CTV budget allocation.

What Walmart and Kroger built, and the verified timeline

The primary source here is Digiday's reporting from 30 June 2026, titled "Streaming is the next frontier for Walmart's, Kroger's ads businesses." The article is a straightforward account of the partnerships and the industry context. The facts below are taken directly from that report and the retailer's own announcements, not from analyst projections.

Walmart Connect, June 2026. Walmart opened Vizio inventory in Walmart Connect via Yahoo's demand-side platform (DSP) through ad-tech provider Magnite. This followed Walmart's 2024 acquisition of Vizio, which gave Walmart access to Vizio's Inscape automatic content recognition (ACR) data: viewing patterns from tens of millions of smart TVs in US households. The Magnite integration makes that inventory accessible to brands buying through Walmart Connect without requiring a direct integration with a trade desk. Separately, Walmart Connect partnered with Google's Display & Video 360 (DV360) to let brands use Walmart audience data to manage and measure YouTube campaigns.

Kroger Precision Marketing, March to June 2026. Kroger launched a YouTube advertising integration via DV360 in March 2026, giving brands access to Kroger shopper audience segments for YouTube targeting and post-campaign purchase attribution. In June 2026, Kroger announced an additional integration allowing advertisers to reach Kroger shoppers within TikTok's advertising platform. Christine Foster, Group Vice President at Kroger Precision Marketing, told Modern Retail: social video "is where people are starting their shopping journeys and thinking about what they're going to do next. This is where a lot of the meal inspiration gets planned, and a lot of health and beauty decisions get made, and we need to be in those spaces."

Albertsons Media Collective, April 2026. Albertsons followed with a YouTube and DV360 integration in April 2026, the third US grocery retailer to connect shopper purchase data to YouTube inventory within a four-month window.

Sean Crawford, Managing Director for North America at SMG, the retail media consultancy, summarised the direction in the same Digiday piece: "The recent wave of streaming partnerships signals that retail media is evolving beyond retailers' owned-and-operated properties into a much broader media ecosystem." That is accurate. What it does not say is that the measurement quality scales linearly with the reach expansion. It does not, for reasons this post covers in section five.

For teams running programmatic buys, these integrations mean the DV360 seat you already operate can now carry Kroger or Walmart audience data on YouTube. That is a workflow change, not a new platform onboarding. The activation path is shorter than it looks.

How the closed loop works: from streaming impression to purchase

The mechanism is straightforward in principle and requires several data joins in practice. The diagram below maps the four steps. Each node is a discrete data event; the connections between them are where match rates, identity resolution quality, and data governance decisions determine whether the loop actually closes.

Step 1: Streaming impression. The brand's ad is served to a CTV device (a Vizio smart TV, a Roku device running a Walmart-connected app, or a YouTube video on a connected screen). The impression is logged with a device identifier and a timestamp.

Step 2: Retailer first-party identity match. The retailer's identity graph maps the CTV device to a known shopper. For Walmart, this uses the Vizio Inscape ACR data joined to the Walmart+ loyalty and account database. The match is deterministic where the shopper is logged into a Walmart account on the same network or device; probabilistic where the join relies on household IP or device graph inference. Match rates vary by retailer and inventory source, and are rarely published. Industry estimates for deterministic RMN identity matching run between 40 and 70 percent of impressions, depending on the audience segment and inventory type.

Step 3: In-store or online purchase. The shopper buys the advertised product, either in a physical store or via the retailer's app or website. The purchase is logged in the retailer's transaction system with the shopper's account ID.

Step 4: Measurement feeds the campaign. The retailer's measurement layer matches the impression log (step 1) to the purchase record (step 3) via the identity graph (step 2). The output is a post-campaign attribution report showing which audience segments saw the ad and subsequently purchased, what the revenue amount was, and often a calculated ROAS figure. This data feeds back into the campaign for audience optimisation in subsequent flights.

The loop is real. The gaps are in steps 2 and 4. Step 2's match rate determines how much of the impression universe the retailer can attribute at all. Step 4's attribution model determines whether "subsequent purchase" means "caused by the ad" or simply "happened after the ad." The latter is a correlation, not a causal claim. Analytics and measurement work sits here: building the holdout methodology that converts matched attribution into incremental proof. Google's DV360 documentation for audience-based buying provides the technical reference for how retailer audience segments are activated in connected video campaigns (blog.google, ads and commerce). Kroger's TikTok integration routes through TikTok for Business's standard advertiser onboarding for audience segment activation, documented in the platform's own resources (TikTok for Business).

What mid-market brands can copy today

The Digiday piece specifically noted that Walmart Connect's Vizio integration targets smaller advertisers: brands without large media teams, including Walmart Marketplace sellers. Martin Kristiseter, CEO of Digital Remedy, told Digiday that the simplified interface means smaller advertisers can launch CTV campaigns and measure performance using Walmart's commerce data without the complexity of a trade desk relationship.

For mid-market brands with product distribution at one of the US retailers mentioned above, the activation path is shorter than it has historically been for CTV. Three practical steps:

1. Confirm shelf presence and account eligibility. The closed-loop measurement only works for products sold at or through the retailer. Kroger Precision Marketing's YouTube and TikTok integrations are most relevant to consumer packaged goods (CPG) brands with product on Kroger-family shelves (Kroger, Fred Meyer, Ralphs, Harris Teeter, and others). Walmart Connect covers Walmart and Sam's Club, plus Walmart Marketplace sellers.
2. Use the DV360 seat you already operate. If your team or agency runs Display & Video 360, the Kroger and Walmart YouTube integrations work inside that seat. This is not a new platform onboarding. Request audience segment access from the retailer's media team and activate through the existing workflow. See how media integration work connects these data layers without rebuilding the entire stack.
3. Build a holdout from the first flight. The first campaign flight should not aim for scale. Aim for a clean holdout design: a matched set of stores or households that do not receive the ad, against which the exposed group's purchase rate can be compared. Without this, the ROAS figure the RMN reports is matched-not-incremental. The holdout makes it incremental.

For brands in Singapore, Malaysia, Australia, and Canada, the honest answer is that retail media CTV at this level of closed-loop measurement is not yet available from domestic retailers. In Singapore, Lazada and Shopee operate growing ad platforms, but neither offers deterministic CTV attribution at Walmart-level depth as of mid-2026. In Australia, Woolworths and Coles have built commerce media offerings and are expanding; CTV integration is nascent. In Canada, Loblaws (PC Optimum loyalty data) has announced retail media expansion, but CTV closed-loop measurement is less mature. In Malaysia, Shopee MY ad products exist with commerce data, but streaming integration is not yet comparable.

The performance marketing framing for this channel is direct: retail media CTV is now a performance channel with a purchase outcome, not a branding channel with a reach outcome. Evaluate it on the same ROAS discipline as paid search. Budget accordingly.

Where the measurement is still soft

The honest version of this post cannot end without itemising what the closed-loop claim does not yet deliver. These are structural gaps, not vendor failures. They apply to the most sophisticated retail media CTV implementations as of mid-2026.

• Match rate opacity. Retailers do not publish their identity match rates. A campaign that reached 10 million CTV impressions may have attributed outcomes from only 4 to 7 million of them, if the match rate is 40 to 70 percent. The unmatched impressions simply drop out of the attribution report. This makes the reported ROAS look better than the actual ROAS on total spend.
• Incrementality is not default. Matching an impression to a purchase is correlation, not causation. The shopper may have purchased regardless of the ad. Sean Crawford of SMG noted in the Digiday piece that he expects only the largest retailers to build fully integrated measurement capabilities. That includes holdout-based incrementality testing, which requires a retailer to give up attributed revenue in the control group to produce a cleaner number. Few do this by default. The IAB's Project Eidos, launched February 2026, is the industry's most structured attempt to address this class of measurement problem at an industry-wide level (IAB, Feb 2026). The June 2026 IAB guidance on AI-powered contextual tools for video is a parallel document that covers what measurement signals are available in streaming environments where deterministic purchase data is absent (IAB, Jun 2026).
• Multi-retailer double counting. A brand selling at both Walmart and Kroger, running campaigns on both RMNs simultaneously, will see purchase events counted in both attribution reports. The same shopper buying a product shows up as a Walmart Connect attributed sale and a Kroger Precision Marketing attributed sale if they happen to be in both RMNs' exposed audiences. Independent measurement via marketing mix modelling (MMM) is the only way to reconcile this. For brands running across multiple RMNs, analytics and measurement infrastructure that sits outside any single retailer's reporting becomes the arbiter.
• APAC lag is 18 to 24 months. None of the closed-loop CTV measurement capabilities described here are currently available from Singapore, Malaysian, Australian, or Canadian retailers at equivalent depth. The US market leads by a structural margin. Planning for it is sensible; budgeting to it now in those markets is premature.
• Data residency and consent. In Singapore (PDPA), Australia (Privacy Act and OAIC enforcement), and Canada (PIPEDA/Law 25 in Quebec), the rules governing how shopper purchase data can be used for advertising are materially different from US norms. Any APAC brand or US brand with APAC audiences that builds a retail media CTV program needs legal sign-off on the data flow between the retailer's transaction environment and the ad platform before activation, not after.

The underlying thesis of this post stands: retail media extending into CTV is a genuine structural shift. A streaming impression can now carry a purchase outcome, without a cookie, using the most defensible data source in advertising: the retailer's own transaction record. The move Walmart and Kroger made in the first half of 2026 is the model every tier-1 retailer globally will replicate.

The caution is equally real: the measurement is better than legacy CTV attribution, but it is not the clean proof-of-incrementality number it is sometimes presented as. Build the holdout. Keep MMM running in parallel. Don't let the retailer's ROAS figure be the only measurement signal in the room.