Retail Media Intel

Executive summary

Retail media's center of gravity is splitting in two directions at once. In the UK, Tesco used its Media Upfront to put influencer-led video across 3,800 in-store screens, mostly in Express short-trip stores, paired with Intelligent Store Orchestration that personalizes by store, time, stock and weather, citing a 47% new-buyer uplift in a Costa trial, and a new self-serve Ad Manager built with Kevel using Clubcard data for AI planning, targeting and measurement. That is creator + store + self-serve AI in one bundle, explicitly trying to make the aisle a measurable lower-funnel channel. In the US, the opposite discipline is taking hold: a Bain survey of 61 network leaders cited via eMarketer shows expected growth concentrating back onsite in search, sponsored product, native AI ads and display, because onsite margins of 70-80% crush offsite near 40% after revenue shares, while offsite social at $1.3bn and in-store face weak advertiser demand on cost and complexity. Meanwhile India shows why proximity still wins budgets: festive ad spend is forecast at INR 60,000-62,000 crore with commerce and retail media at INR 11,500 crore, up 27% year-on-year and about 19% of festive spend, and quick-commerce advertising expected to surge 50% to INR 2,200 crore. The cross-domain bridge is agentic AI and identity. Capgemini Invent frames the shift as from selling ads to shaping decisions as agents recommend rather than rank sponsored listings, Ulta Beauty CMO Kelly Mahoney argues AI discovery via ChatGPT and Gemini is driving more confident store visits, not replacing them, with NielsenIQ showing 57% of Gen Alpha AI users visit stores to browse versus 36% of non-users, powered by creator and associate-led Ulta Beauties content. THG Commerce linking with Criteo on a Gemini Enterprise-based AI shopping assistant points the same way: assistants monetized as retail media. Underneath, commerce infrastructure is becoming media infrastructure. Google's Merchant Center loyalty_program attribute and MemberProgram schema plus Tesco Clubcard-style single-ID loyalty are positioned as the fuel linking merchandising, Google surfaces and retail media targeting, while Teads and Koddi opening Sponsored Product and Display on Wolt Ads, Gopuff UK and Hopper via OpenRTB, Best Buy Ads selling LG TV inventory paired with shopper plus device data, and Google India pushing brandformance with 1.6x higher ROI versus paid social for Lenovo India, Ajio and Zepto all show feeds, loyalty, CTV and programmatic converging on closed-loop commerce.

Market dynamics

The structural force is fragmentation versus infrastructure. IAB's Collin Colburn and the IAB UK roundtable with Criteo, PubMatic, TripleLift, The Trade Desk, Moloco and Zitcha both name the same problem: 100-plus networks split across onsite, offsite and in-store with no shared buying or measurement. The consensus fix is interoperability, persistent identity, middleware to unify buying, MCP links between buy-side AI and retailer networks, and measurement beyond ROAS to incrementality and lifetime value. DMEXCO's Ad Context Protocol backed by 100-plus partners for agentic CTV buying is the same instinct in TV. Budget flow is ruthlessly pragmatic. High-margin onsite search is subsidizing experimentation elsewhere, quick-commerce and marketplace media are outgrowing total ad spend on closeness to transaction in India, and brand-plus-performance AI campaigns via Performance Max and AI Max are being sold as the escape from a CAC valley. Tech is lowering the cost of entry while raising the bar for proof: Kevel-powered self-serve, Koddi-powered OpenRTB access where retailers keep pricing and inventory control, and AI orchestration of creative by context all reduce integration friction, but advertisers are demanding incrementality over attribution and exposed-user testing over panels before they fund in-store screens or offsite extensions at scale.

Key developments

Strategic implications

Builders should prioritize onsite excellence first, then portable proof. Fund roadmaps from 70-80% margin search, sponsored product and AI-native onsite formats, and make every offsite or in-store extension carry incrementality, not ROAS, with store, time, stock and weather context like Tesco's orchestration and footfall plus in-stock closure from hyperlocal and quick-commerce zones. If in-store creator cannot prove new buyers like the 47% Costa example under controlled exposure, it will not scale beyond pilots. The cross-domain wedge is identity plus agent-readiness plus programmatic access. Unify loyalty ID, Merchant Center feeds with member pricing and reviews split correctly between Seller Ratings and product-level Shopping snippets, and retail media audiences so one purchase graph powers merchandising, Google free listings and onsite targeting. In parallel, expose inventory via OpenRTB and MCP-compatible interfaces like the Koddi-Teads pattern so buy-side AI and middleware can buy without custom work, and author product content for both humans and agents with prompt-informed copy, associate and creator proof, and structured availability and pricing. Treat CTV and quick-commerce as the two offsite bets with real transaction proximity. Pair shopper data with TV OS inventory for incremental reach rather than generic extension, and treat hyperlocal radius, polygon and delivery-zone targeting tied to local inventory ads as the measurement bridge between proximity media and commerce outcomes, reinforced by brandformance balancing to avoid pure performance CAC traps highlighted by Google India.

Competitive landscape

Scale retailers with data plus screens are setting pace: Tesco with Clubcard plus 3,800 screens and Kevel self-serve, Walmart and Best Buy Ads extending shopper data into CTV via LG and VIZIO, and quick-commerce leaders like Zepto, Wolt and Gopuff monetizing closeness to basket. Google remains the cross-domain gravity well across Gemini, Performance Max and AI Max, Merchant Center feeds and loyalty schema, and YouTube, while infrastructure players Criteo, Koddi, Kevel, Zitcha, PubMatic and The Trade Desk compete to be the interoperability layer. Agencies like Brandtech with 500 specialists in Brandtech Data are productizing AI-ready data to tie exposure to business outcomes, leaving point-solution networks without incrementality proof most exposed.

Action items

Trend lines

Contradictions

• Calls to retreat to onsite profitability and incrementality clash with continued CTV/offsite expansion via Best Buy Ads, ShowHeroes, and omnichannel pushes.
• Push for open standards and programmatic unification contradicts walled-garden enforcement such as Meta banning TikTok ads in seven markets.
• Creator authenticity narrative conflicts with AI-automation scale narrative driving the same in-store screens and discovery surfaces.

Blind Spots

• No store-level incrementality or sales-lift results for creator in-store content - pilots and intent only, no closed-loop proof yet.
• Limited insight into costs, margins, and pricing for onsite versus offsite to validate the profitability retreat thesis.
• Geographic skew to US/UK with only thin signals from India festive wave and Indonesia; no privacy, regulatory, or small-retailer viability assessment.

Connections

• Clubcard-style loyalty data is becoming joint fuel for in-store screen targeting and AI-assistant personalization, linking retail-media measurement to commerce retention.
• Quick-commerce programmatic slots for Wolt and Gopuff via Koddi/Teads turn delivery apps into onsite-like inventory, connecting hyperlocal commerce fulfillment to retail-media standardization.
• AI discovery via ChatGPT, Gemini, and THG/Criteo assistants collapses Shopping ads, product feeds/reviews, and retail-media search into one conversational surface, forcing joint feed optimization for commerce and media.

Article index