What eCommerce brands get wrong about retail media measurement

By Shani Rosenfelder

TL;DR

Retail media networks measure within their own walls — that’s by design, not a flaw, but it means network-reported ROAS was never built to answer cross-channel questions
A 63% fluctuation in how ROAS is calculated network to network isn’t a data quality problem, it’s a methodology problem no amount of in-network optimization can fix
Brands working across multiple retail media networks pay a fragmentation tax: each network becomes its own source of truth, and none of them reconcile
Independent measurement doesn’t replace what retail media networks report; it gives brands the layer to interpret, reconcile, and act on it
The same signal infrastructure that solved mobile measurement’s hardest signal governance challenges is what retail media measurement needs to catch up to

Mind the ROAS gap

Here are two numbers worth sitting with. A 2026 study by Albertsons, Northwestern University, and Ovative Group found a 63% fluctuation in how ROAS is calculated across retail media networks — 11 variables, same campaign, wildly different figures depending on whose methodology you use. Separately, Skai and Stratably found that only 15% of brands report strong measurement confidence in retail media.

Those two numbers aren’t unrelated. The confidence gap is a direct consequence of the methodology gap. When the numbers don’t reconcile, you stop trusting them.

Retail media’s closed-loop measurement isn’t broken. But it is incomplete by design — and the brands scaling budgets across networks without an independent measurement layer are making portfolio decisions on data that was never built to support them.

Why retail media measurement is built the way it is

Retail media networks measure within their own walls because that’s how closed-loop attribution works — a structural feature, not a design flaw. Each network has its own first-party customer data, its own attribution windows, and its own conversion logic. It can only see what happens inside its own ecosystem.

This is genuinely valuable. Closed-loop measurement gives eCommerce brands something most channels can’t: real purchase data tied to ad exposure. For a brand trying to understand whether a sponsored product placement drove a sale, that signal is real and meaningful.

The problem starts the moment you try to compare what one network reports to what another does — or use either figure to make a cross-channel budget call. That’s a question closed-loop measurement was never designed to answer.

What the 63% ROAS fluctuation actually tells you

The 63% fluctuation in ROAS calculations isn’t noise. It’s the direct result of 11 specific methodology variables that each network controls independently:

Attribution window length — does a sale count if it happened seven days after ad exposure, or only one?
How view-through conversions are weighted versus click-through
How incremental sales are defined and how the baseline is set
Whether halo effects on non-advertised products are included
How new-to-brand customers are identified and counted

Each is a legitimate methodological choice. Each produces a different number.

For an eCommerce brand running a seasonal campaign — say, a push ahead of a major shopping period — the practical implication is stark. The same SKU, promoted across two networks simultaneously, will produce different ROAS figures not because one network outperformed the other, but because they counted differently. A 4x return on one network and a 2.8x on another might reflect identical real-world performance viewed through different lenses.

This is the core mistake. When eCommerce marketers see diverging ROAS figures and conclude that one network is simply performing better, they’re often looking at a methodology difference, not a performance difference. Budget decisions made on that basis aren’t just unreliable — they’re structurally misleading.

What happens when every network becomes its own source of truth

When eCommerce brands scale across multiple retail media networks — and most do; brands worked with an average of six networks in 2025, according to The Current — the fragmentation tax compounds with every network added. Six networks means:

Six attribution logics, each counting conversions differently
Six conversion window definitions, none of them aligned
Six dashboards producing confident-looking numbers that can’t be reconciled

The specific mistake this creates is using network-reported ROAS to rank networks against each other. It feels like a reasonable way to allocate budget — run the same product across networks, compare the returns, shift spend toward the winner. But when each network calculates ROAS differently, you’re not comparing performance levels. You’re comparing measurement frameworks and mistaking the difference for signal.

For brands under pressure to defend retail media investment to finance teams and leadership, this isn’t an analytics inconvenience — it’s a strategic liability. The problem isn’t missing data. It’s irreconcilable data. And adding more networks without a reconciliation layer doesn’t scale insight. It scales fragmentation.

What independent measurement actually changes

Independent measurement doesn’t replace what retail media networks report. It provides the layer above them — where the same attribution logic, the same signal infrastructure, and the same conversion data applies across every network simultaneously.

With that layer in place, network-reported figures become inputs rather than verdicts. Each network’s dashboard stays useful for what it was built to do — reporting on its own inventory. But the budget decisions, the cross-network comparisons, the question of what’s actually driving eCommerce conversion — those get answered somewhere the networks can’t reach.

This is where retail media data stops being a siloed report and starts being an actionable signal — connected to the rest of the channel mix, interpreted through the same logic as search, social, and every other channel in the eCommerce stack. When the same signal infrastructure runs across channels and conversion data flows back to every network from one place, eCommerce brands stop maximizing individual networks and start maximizing their business.

The measurement standard that makes this possible already exists. Mobile performance marketing built it — under the toughest signal governance conditions in the industry, navigating platform fragmentation, privacy constraints, and sophisticated fraud. That standard is now what every channel needs, retail media included.

That infrastructure is also under pressure. When Publicis acquired LiveRamp — following WPP’s takeover of InfoSum — the category lost its last major independent players. The same conflict-of-interest risk brands recognize in self-attributing networks now applies to the measurement layer itself

Final thoughts

Retail media’s measurement isn’t broken. But it was built to serve each network’s reporting needs — not an eCommerce brand’s cross-channel decision-making needs. That gap doesn’t close from inside any individual network’s dashboard.

The brands that will make the most confident retail media budget decisions aren’t the ones with the most network dashboards. They’re the ones with a single measurement layer that sits above all of them — where the same logic applies everywhere, the data reconciles, and portfolio-level decisions have a reliable foundation.

Omnichannel measurement confidence in retail media isn’t a function of better network reporting. It’s a function of independent measurement — and the signal infrastructure to support it already exists.