Retail media ROAS: five questions before comparing networks

One retail media network reports a 6× return on ad spend. Another reports 4×. Moving budget toward the higher number looks straightforward until you discover that one includes sales after ad views and the other counts only sales after clicks.

Retail media uses a retailer’s advertising inventory, shopping data, or both to reach customers and measure results. Access to purchase data can make reporting useful, but it does not make every reported return comparable.

IAB Europe’s Commerce Media Measurement Standards V2.1, published in January 2026, include clearer gross and net sales definitions. That is a useful starting point for a comparison sheet. It is still necessary to establish what each actual report contains.

1. Which sales receive credit?

Ask which interactions qualify for attribution, the period after each interaction in which a sale can receive credit, and how competing interactions are handled. Attribution is the rule for assigning credit; it is not proof that the ad caused the purchase.

A seven-day click window and a fourteen-day view window answer different questions. Do not combine them under one unlabeled ROAS column.

Record whether the report uses the interaction date or conversion date. A campaign ending on Sunday can still receive later sales credit if its attribution window remains open. Set a date for the final readout rather than treating the first export as settled.

2. Which products and customers are included?

Separate purchases of the advertised product from purchases of other products in the same brand. If a report includes broader brand sales, say so.

For “new-to-brand” results, ask how new is defined, how far back the network checks purchase history, and what happens when the shopper cannot be recognized. A customer may be new within one retailer’s records while having bought the brand elsewhere.

Match the question to the campaign objective. A launch intended to recruit customers needs a different interpretation from a campaign promoting a refill to existing buyers.

3. Are sales gross or net?

Returns, cancellations, discounts, and other adjustments can change the number used as revenue. Establish which are included and when the report is updated.

Consider two illustrative reports with $10,000 of ad spend each:

Report	Attributed sales shown initially	Later returns and cancellations	Sales after those adjustments
Network A	$60,000	$12,000	$48,000
Network B	$50,000	$2,000	$48,000

The initial ROAS figures are 6× and 5×. Using the adjusted sales in this example gives 4.8× for both. That does not make the campaigns equivalent: their attribution rules, customer mix, and margins may still differ.

Use the ROAS calculator once the revenue definition is consistent.

4. What is in the cost denominator?

Determine whether ad spend includes only media or also platform, data, creative, and service fees. Keep the reported platform ROAS if it is useful, but label it clearly.

For a business comparison, calculate a second version using an agreed cost basis. Do not quietly add fees to one network and leave them out of another. Revenue divided by advertising cost also remains different from profit, which depends on product and other business costs.

5. Is the decision about credit or additional sales?

A strong attributed ROAS can coexist with limited incremental sales if many shoppers would have purchased anyway. Incrementality asks what changed because the advertising ran.

The IAB and IAB Europe incremental measurement guidelines discuss methods for estimating that additional impact. A causal question needs an appropriate study; relabeling attributed revenue will not answer it.

Make the differences visible

Keep a comparison sheet beside the dashboard with the attribution windows, qualifying interactions, product scope, customer definition, revenue adjustments, cost basis, and reporting maturity date.

If the networks cannot provide matching definitions, present the results side by side with those differences visible. You can still make a decision, but the decision should acknowledge what the numbers can and cannot establish.