A closed loop is one of the most consequential ideas in performance marketing right now. It means the company that served the ad can also see whether it led to a sale. Instead of counting clicks or modelled conversions, a retailer or platform matches its ad-exposure records against actual purchase records from loyalty cards, logins, hashed emails or card data.
Why it matters
Most advertising is seen in one place and paid for at a till elsewhere. A retail media network changes that. A shopper sees a sponsored product listing, later scans a loyalty card at checkout, and the network joins the two logs. The result is a return on ad spend figure based on transactions, not proxies.
The Media Rating Council formalised a narrower definition — the closed-loop auction — in January 2026: an auction where one platform hosts every function, usually without OpenRTB. The MRC says the vast majority of auction-based spend now trades this way across search, social and retail.
One campaign, three ROAS numbers
PPC Land’s example shows why the same number can mean very different things. A brand spends $50,000, and identified shoppers who saw the ads buy $150,000 of the promoted item within 14 days: that is a 3.0 ROAS. If three quarters of sales are identifiable and unidentified shoppers are assumed to behave the same way, attributed sales become $200,000 and ROAS becomes 4.0. A holdout test showing $120,000 would have happened anyway drops incremental ROAS to 1.6. All three can be labelled closed-loop.
The IAB and MRC retail media measurement guidelines, published in January 2024, require lookback windows to be disclosed before a campaign runs; common windows run 3, 7, 14, 28 or 30 days.
Where the loop leaks
The uncomfortable part is that the party measuring the result often sold the ad. The Trade Desk CEO Jeff Green argued in August 2025 that a platform’s bias toward its own inventory creates conflicts that prevent objective buying. A vendor study of 150,000 campaigns and $350 million in spend, published in June 2026, estimated that siloed attribution misses 36% to 53% of campaign impact — rising to 67% to 80% for off-site video — because shoppers exposed at one retailer buy at another.
Auction transparency adds another concern. The MRC standards require auctioneers to disclose the auction type, report reserve prices before and after each auction and apply one floor to every buyer. But compliance is voluntary, and Amazon and Google were not listed among participants. In August 2026 the FTC and 22 states sued Amazon over about $20 billion in alleged hidden ad surcharges. No claim has been tested in court.
What to check before scaling
Closed-loop measurement is genuinely useful, especially in retail media and search. Just don’t treat it as a verdict. Before trusting a reported ROAS, ask:
- Is this attributed ROAS or incremental ROAS?
- How much of the result is modelled or extrapolated?
- Which tills does the loop actually see?
- What auction details is the platform disclosing?
Retailers are now pushing shopper data into third-party DSPs — Walmart Connect into DV360 and Yahoo DSP, Kroger into Google’s Commerce Media Suite. That makes the measurement questions more urgent, not less.
Source: PPC Land
