Retail media was built around deterministic audiences that could sit in a segment for months. That model is giving way to something faster: real-time customer context, where consent, identity and a shopper’s latest behavior can shape the next ad, offer or product recommendation in milliseconds.

Jon Flugstad returns to The Middlemen to explain why the identity infrastructure that helped launch the first generation of retail media networks is starting to become a constraint. Renting identity from a third party once made sense, but rising costs, slow audience updates and dependency on outside match tables make it harder for retailers to act on their own data in real time.

The conversation explores a broader definition of context. It is no longer just the content on a page or a shopper’s historical purchases. Customer context combines identity, consent and a live sequence of behaviors to help retailers predict what someone is likely to do next. Jon shares examples in which real-time behavioral signals produced substantial improvements in ad performance and conversion rates.

Tom, Scott and Jon also debate whether paid and organic results should be optimized together, why each still requires a different machine-learning model, and what retailers risk when they allow valuable data and inventory to flow too freely into programmatic markets.

The episode then turns to offsite retail media. Direct connections to platforms such as Meta, Google and TikTok can give retailers more control over the signals they share while making campaigns easier for brands to buy. Better identity matching, richer first-party signals and faster activation can help retailers grow offsite without building enormous managed-service teams.

At the center of the discussion is a strategic question:

‍