Digital advertising has an identity crisis — and it’s getting worse. The tools that the industry relied on for two decades to identify users, track behaviour, and deliver relevant advertising are being systematically dismantled. Retail media is the most compelling structural solution to this problem.
What the Industry Lost
For years, digital advertising ran on third-party cookies and device identifiers. Apple’s App Tracking Transparency framework required explicit user opt-in for cross-app tracking on iOS, materially degrading mobile advertising signal quality. Then Google announced the phase-out of third-party cookies in Chrome — which holds roughly 65% of global browser market share. The result: advertisers are flying increasingly blind. Targeting precision is down. Attribution is harder. The ROAS numbers that justified digital budgets for years are becoming harder to defend.
Why Retail Data Is Different
Retail media operates on an entirely different data foundation. When a consumer logs into an e-commerce platform, their identity is known — not inferred. Their purchase history is real. Their browsing behaviour is observed directly. This is first-party data at its best: collected in the context of a commercial relationship, consented to by the user, and actionable at the moment of purchase intent. No cookie matching. No signal degradation from privacy changes. Industry data suggests average return on ad spend (ROAS) for retail media campaigns in the range of 20–40x — numbers that represent a step-change improvement over other digital channels.
Off-Site: Taking Retail Data Beyond the Retailer’s Website
The second critical capability is off-site activation. Using a retailer’s first-party data matched against the open web through privacy-safe clean rooms, brands can serve targeted advertising across publisher inventory, social platforms, and connected TV — without relying on third-party cookies. This effectively gives advertisers a way to do what cookies used to do, but better: using verified purchase data rather than proxied behavioural signals, with closed-loop attribution back to actual transactions.