In this Retail Media Summit UK session, News UK’s Charlie Celino and Permutive’s Tom Shapland tackled one of retail media’s most urgent questions: what happens when identity collapses? As cookie deprecation accelerates and match-rate obsession continues, the pair argued that retailers and advertisers need to evolve their thinking beyond IDs toward richer, more scalable signals that live across the open web.
The Bottom Line
As cookies disappear and match rates flatten, the open web becomes retail media’s biggest untapped opportunity. News UK and Permutive showed how contextual intelligence, sentiment analysis, and predictive modelling can enrich retailer first-party data, powering smarter off-site activations, broader reach, and deeper customer insight in a privacy-first world.
Key Takeaways
IDs Aren’t Going Away But We Need More Than an Arms Race
Celino didn’t mince words: the industry has become “a little bit obsessed with IDs.” They’re valuable, but they’re not the whole story and, in many cases, they’re incomplete or misleading.
“We act like we need to capture everyone and everything… but we have to be aware of the limitations.” — Charlie Celino
Households share loyalty cards, legislation is tightening, and deterministic matching only covers a fraction of real buying behaviour. To evolve, retail media has to pair ID-based precision with signals that describe context, sentiment, and intent, not just a single user’s login.
The Open Web Is Full of Signals Retailers Aren’t Using Yet
With only ~30% of the open web reachable via cookies or IDs, publishers like News UK have leaned heavily into contextual and behavioural intelligence. In Celino’s words, context is “not just headline-level anymore.”
“It’s not just contextual signals — it’s sentiment, emotion, scroll depth, format type… all of it informs market decisions.”
With advancements in LLMS and video understanding, publishers can now classify content more deeply and read how users engage with it. That creates a powerful complement to retail first-party data, especially in consideration moments where consumers turn to reviews, lifestyle content, news, and cultural moments to influence choice.
Predictive Modeling Unlocks Scale Without Sacrificing Privacy
When match rates fall, predictive modelling steps in. But Shapland and Celino were clear: this isn’t guesswork — it’s the same modelling that powers Meta, Google, and major platforms post-ATT.
“The same probabilistic modelling that Meta uses after ATT is what the open web needs now.” — Tom Shapland
Start with rich seed data, then build high-quality probabilistic lookalikes (Celino cited 77–88% accuracy thresholds). This opens up category-level insight (e.g., deodorant buyers across retailers) and allows retailers to add “colour” to their understanding of customers — behaviour, interests, journeys — not just transactions.
Quote to Remember
“Everyone is an individual. A postcode won’t tell you who they are. What they read, where they read it, when they engage — that’s where the real colour comes from.” — Charlie Celino
