Closed-Loop Attribution

A measurement approach that connects every marketing touchpoint — ad impressions, clicks, emails — to actual downstream sales outcomes, "closing the loop" between ad event and purchase. Unlike models that stop at the click, it follows the customer journey all the way through to the transaction and feeds those outcomes back into reporting and optimisation.

In Retail Media specifically, the loop closes because both ends of the shopper journey — the ad event and the sales event — sit inside the same data environment, owned by the retailer, and resolved against the same identity graph (loyalty ID, logged-in account, or payment instrument).

How it works

The canonical data flow (as described by Topsort and AppsFlyer):

Ad event captured — impression or click tied to campaign ID, product ID, placement ID, and a session/user identifier.
Shopper journey continues — product views, add-to-cart, browsing signals recorded.
Purchase event fires — order value, SKU, quantity, timestamp recorded.
Matching — the purchase event is matched back to the prior ad event if it falls within the Attribution Window (typically 7–14 days last-click for onsite sponsored products; 30 days view-through for display).
Revenue attributed — credit assigned per the platform's attribution rules (last-touch, multi-touch, SKU-level, or basket-level).
Reporting closed — attributed ROAS = attributed revenue ÷ ad spend surfaced in the advertiser dashboard.

Sources: Topsort (June 2026); AppsFlyer glossary

Attribution sub-models used inside closed-loop retail media

Click-through attribution — most common for sponsored products
View-through attribution — display and video; higher over-crediting risk
Last-touch attribution — simple, widely used, undervalues upper funnel
Multi-touch attribution — distributes credit; complex to implement
SKU-level attribution — promoted product only
Basket-level attribution — full order value including non-promoted items

Source: Topsort (June 2026)

How it differs from other models

Model	What it answers	Key weakness
Last-click attribution	Which channel got the final click?	Over-values bottom-funnel capture
Multi-touch attribution (MTA)	How to distribute credit across touchpoints?	Degrades without cookies/IDs; walled-garden blind spots
Media Mix Modeling (MMM)	What is long-run channel contribution?	Slow (weeks/months lag); not actionable at campaign level
Closed-loop attribution	Which sales are connected to ads?	Proves correlation, not causation; self-attribution risk
Incrementality	Which sales were caused by ads?	Requires large sample; takes time to run

Critical distinction: Closed-loop attribution and Incrementality answer different questions. Attribution shows correlation; incrementality proves causal lift. The emerging industry consensus is both are needed — MMM gives the strategic telescope; MTA gives the execution microscope; closed-loop gives the real-time performance readout.

Sources: Haus.io (2025); Osmos (2026)

Key platforms (as-of 2026)

Amazon Ads

Amazon Marketing Cloud (AMC): SQL-based clean room; lookback window extended from 13 to 25 months at unBoxed 2025.
Amazon Attribution: tag-based measurement product for non-Amazon channels feeding back to Amazon purchase data.

Source: Osmos (2026)

Walmart Connect

Multi-Touch Attribution model unified across onsite, offsite, and in-club.
Walmart Luminate clean room rebranded to Scintilla (post-Vizio acquisition); Vizio integration enables unified CTV + in-store + ecommerce measurement loop.

Source: Osmos (2026)

Instacart

Launched Data Hub on January 6, 2026: a privacy-safe clean room allowing CPG brands to join their own first-party data with Instacart grocery purchase signals for audience building and flexible attribution.
2026 sponsored products ROAS benchmark: 4.8–6.7x; conversion rates 18–25%, the highest of any retail media platform. (as-of 2026)

Source: Osmos (2026)

Kroger Precision Marketing

95% of Kroger grocery sales are linked to loyalty accounts (as-of 2026), enabling a genuinely deterministic closed loop — one of the strongest identity graph positions in retail.

Source: MetaRouter

Albertsons Media Collective

Closed-loop capability built on matching loyalty card purchase data to ad impressions — giving CPG brands deterministic proof of sales lift, not modelled estimates.
Co-published landmark study with Ovative Group and Northwestern Kellogg School (see Limitations below).

Source: YouTube — Liz Roche, Shoptalk 2025

Grocery TV (in-store)

Launched first closed-loop measurement solution for a national in-store retail media network, April 2025. Hy-Vee was the first retailer to deploy it.
Early benchmark: average 4.7x ROAS for CPG advertisers. (as-of April 2025)

Source: PR Newswire (April 2025)

Benchmarks (as-of 2026)

US retail media ad spend 2026: $71.09B, +17.8% year-over-year. (as-of 2026) Source: Osmos
Cross-industry retail media ROAS average: 6.1x, maintained for five consecutive quarters. (as-of 2026) Source: Osmos
86% of commerce media decision-makers in North America and Europe rank strengthening measurement and attribution as a high or critical priority heading into 2026. (as-of 2026) Source: Advertising Week
Only 12% of commerce teams have reached full-funnel closed-loop attribution maturity. (as-of 2026) Source: Osmos
75% of advertisers cite Incrementality as their single biggest measurement challenge. (as-of 2026) Source: Osmos
80% of retail media ad types use last-touch attribution as their default measurement. Source: Incremental (150-campaign analysis)
eMarketer revised its retail media CTV ad spend forecast for 2024 from $1.5B to ~$3.6B, driven by retail closed-loop data flowing into TV buying platforms. (as-of 2024) Source: YouTube — The Trade Desk Forward '24

What practitioners report

The attribution ≠ causation gap

The most consistent practitioner tension is the conflation of attribution (which ads connected to sales) with Incrementality (which ads caused sales that wouldn't have happened anyway). Retail media's closed loop proves correlation inside the walled garden; it does not prove causation.

Liz Roche, SVP Albertsons Media Collective (Shoptalk 2025): "Closed-loop capability gives CPG brands deterministic proof of sales lift — but frames Incrementality, not ROAS, as the correct north-star metric for retail media measurement." Source: YouTube

The Albertsons/Ovative Group/Northwestern Kellogg study of 42 real onsite display campaigns found iROAS can vary by 6.5x — and flip results in 83% of campaigns — based solely on methodology (test/control group filtering, matching method, data features used, and incremental revenue calculation method). Source: Osmos (2026)

Metrics aren't fit for purpose

Liz Roche, VP Media and Measurement, Albertsons: "The metrics that we have today aren't sufficient. They don't accurately reflect what's going on in the business." Source: The Current (April 2026)

Cara Pratt, Circana President: "It is near impossible for a brand marketer to look with consistency and continuity about how dollars are returning business results...when you look across the magnitude of retailers in the ecosystem." Source: The Current (April 2026)

ROAS varies 63% across networks due to methodology differences (Albertsons/Ovative Group/Northwestern Kellogg study). Source: The Current (April 2026)

Elizabeth Marsten, VP Commerce Media at Tinuiti: "I kind of want to set the whole thing on fire." Multiple brands at Tinuiti are running MMM from one provider alongside two separate third-party measurement companies simultaneously — because no single source is trusted. Her advice: abandon ROAS and iROAS as cross-network comparators; focus on top-line sales and new-to-brand metrics. Source: The Current (April 2026)

The CMO accountability shift

Marion Hargett, U.S. CRO at MiQ (August 2025): "Closed-loop attribution is the primary accountability test for brand CMOs — connecting media spend to actual sales is now the entry requirement for programmatic and retail media buys." AI is collapsing the lag between ad exposure and attribution insight: what previously took weeks now happens in near real time via MiQ's trading agent. Source: Beet.TV (August 2025)

Limitations

1. Attribution does not equal causation (the incrementality gap)

Closed-loop attribution connects sales to ads but cannot prove the ad caused the sale. A shopper may have purchased anyway. The Albertsons/Ovative/Kellogg study: iROAS varies by 6.5x and flips results in 83% of campaigns based solely on methodology. Source: Osmos (2026)

2. Self-attribution bias (the scorekeeper problem)

Retail media networks act as both media seller and measurement authority — they set attribution windows, conversion logic, view-through rules, and reporting frameworks. Each platform claims credit independently with no neutral arbitration across channels. Google's PMax and Meta's Advantage+ are specifically cited as prone to selecting in-market users who are already close to buying, then claiming credit when they convert. Sources: Taboola; Nielsen (2025)

3. Walled garden opacity

Advertisers receive only aggregated, platform-curated metrics with no access to raw impression-level or user-level data. Roughly four-in-ten marketers report that lack of data from walled gardens is a measurement problem. Independent verification of campaign performance is structurally impossible. Sources: DoubleVerify; Digiday

4. Clean rooms are pay-to-play

Access to retailer clean rooms — the main tool for independent cross-channel verification — is typically locked behind joint business plans. Measurement accuracy is therefore a function of how much you spend, not how sophisticated your methodology is. Source: The Current (April 2026)

5. Double-counting across campaigns

If a buyer runs Sponsored Products and Sponsored Brands simultaneously on Amazon and a user clicks both before purchase, the sale appears in both campaign reports. Practitioners who sum attributed revenue across ad types routinely overstate actual revenue by 20–40%. Source: Feedvisor

6. The in-store gap

Around 90% of purchases at major grocery retailers still happen in-store (as-of 2026). If 83% of retail revenue is in-store and 77% of retail media spend is measured only on ecommerce, the "closed loop" covers a minority of the commercial activity it claims to attribute. Source: MetaRouter

7. Scale-incrementality tension

Smaller retailers lack the transaction volume to run statistically valid holdout experiments, pushing them toward modelled attribution rather than deterministic closed-loop. Source: YouTube — Epsilon/Austin Leonard (May 2025)

8. Privacy constraints tightening

Cross-device tracking is less complete, consumer consent has greater influence over data availability, and browsers restrict identifiers. This degrades the user-level attribution foundations that closed-loop models depend on. Source: Advertising Week

Recent developments (2025-2026)

February 2025: IAB Tech Lab finalized ADMaP 1.0 — attribution data matching protocol for privacy-safe retail media measurement standardization.
April 2025: Grocery TV launched first closed-loop measurement solution for in-store retail media; Hy-Vee first to deploy. Source: PR Newswire
July 2025: IAB Tech Lab finalized PAIR 1.1 — common protocol for privacy-safe audience activation and data sharing between retailers and brands. Source: Grocery Dive
Amazon unBoxed 2025: AMC lookback window extended from 13 to 25 months. Source: Osmos
January 6, 2026: Instacart launched Data Hub (privacy-safe clean room). Source: Osmos
2026: Retail media + CTV convergence — Walmart/Vizio and Amazon Streaming TV leading the extension of closed-loop measurement to upper-funnel video. Source: eMarketer
Standardization gap persists: IAB standards were published January 2024; as of 2026 no one is operationalizing them at scale. LiveRamp VP Christine Grammier: "The tech is there; [some retailers are] just not interested." Source: Grocery Dive

Standardization progress: IAB Tech Lab reports ADMaP 1.0 and PAIR 1.1 finalized in 2025, framing this as meaningful progress. LiveRamp (via Grocery Dive) says "no one is operationalizing them — some retailers are just not interested." Both sources dated 2025-2026.

Key terms

Term	Meaning
Attribution window	The time period after an ad event during which a subsequent purchase is credited to that ad
Last-touch attribution	All credit assigned to the final touchpoint before purchase
View-through attribution	Credit assigned based on an ad impression even if no click occurred
SKU-level attribution	Only the revenue from the specific advertised product is counted
Basket-level attribution	Full order value (including non-advertised items) credited to the campaign
Clean room	A privacy-safe environment where two parties can match data sets without exposing raw user-level data
iROAS	Incremental ROAS — ROAS calculated only on sales caused by the ad, not correlated with it
Self-attribution bias	When the platform that sells the media also measures its own performance without independent verification
ADMaP	IAB Tech Lab's Attribution Data Matching Protocol (v1.0, February 2025)
PAIR	IAB Tech Lab's Publisher Advertiser Identity Reconciliation protocol (v1.1, July 2025)