Practical methods, not a client performance claim. Any worked example is fictional. Actions require the relevant evidence, authority and spending boundary.
A strong ROAS can coexist with weak economics
A retail media campaign can report attractive attributed sales while discounts, returns, product costs and fulfilment consume the margin. Before increasing spend, I would reconcile what the numerator includes and what the business must pay to deliver those orders.
My retail media background includes contract work with an e-commerce company on Shopee and momo campaigns, product listings, product pages and promotion coordination. This note explains a review method. The financial example is fictional and is not a claim about that engagement.
Define the revenue and order population
Write whether the ad report uses gross order value, paid orders, net product sales or another value. Check tax, shipping revenue, discounts, cancellations, refunds, attribution window, timezone and currency. Keep new versus returning customers and product mix visible when those fields are available.
Match the ad report to commerce records using authorized order identifiers. Do not simply add revenue attributed by several advertising platforms: more than one can claim credit for the same order. An attributed order is not proof that the ad created an incremental sale.
Shopify distinguishes sales reporting from payment movement, and explains that discounts and refunds affect net sales and reported margins. These are reference definitions, not a claim of Shopify operating experience or a substitute for Shopee and momo settlement definitions. Sales reports and profit reports.
Ask what remains after the order is fulfilled
Build an agreed contribution view with the person responsible for the commerce records. Start with net revenue on a stated basis, then subtract product cost and the variable costs required to obtain and fulfil that revenue. Depending on the business, these can include marketplace or payment fees, packing, fulfilment, shipping subsidies and return handling. Avoid subtracting a cost again if it is already netted out.
Contribution after media = net revenue − agreed variable costs excluding media − media spend. State which costs are included. This is not automatically net profit: fixed overhead, taxes, financing and other costs may remain. Sales, contribution and cash settlement answer different questions.
If the contribution margin before advertising is stable and calculated on the same net-revenue basis, an illustrative break-even media ROAS is 1 ÷ contribution margin rate. A blended margin can hide a losing SKU. Changing discount, return rate or product mix changes the threshold; there is no universal “good ROAS.”
Worked example: a reported 4× ROAS
Fictional arithmetic example. All USD amounts are invented to show the distinction. They are not James's results or recommended targets.
| Media spend | $1,000 |
|---|---|
| Reported attributed gross product sales | $4,000 → reported ROAS 4.0× |
| Discounts and refunds | $400 + $400 |
| Net product revenue | $3,200, excluding tax and shipping revenue |
| Product cost retained after returns | $1,800 |
| Marketplace and payment fees | $250 |
| Fulfilment, shipping subsidy and return handling | $350 |
| Contribution before media | $800: $3,200 − $1,800 − $250 − $350 |
| Contribution after media | −$200: $800 − $1,000 |
Here the before-media contribution margin is 25% of net revenue. Its break-even media ROAS would be 4.0× on that same net basis, while the actual net-revenue ROAS is only 3.2×. The headline 4.0× uses a different numerator and hides the shortfall. If refunds or costs remain unresolved, label even this calculation provisional.
Check stock and the promotion before blaming the ad
A campaign report should be read alongside product availability. Check whether the advertised variant is in stock, the promotional price actually applies, shipping conditions match the offer and the product page works on the main device. Compare the page the customer sees with the ad's promise.
A conversion-rate decline during a promotion can come from a price change, unavailable variants or checkout friction as well as traffic mix. Split the review by SKU and comparable period before drawing a conclusion. Avoid changing the bid, creative, discount and product page together: the resulting movement will be difficult to interpret.
For the next bounded test, change one controllable factor and keep the other conditions stable where possible. If a product cannot be fulfilled or a promotion is misleading, correcting that condition takes priority over an optimization experiment. Record the operational interruption so it is not mistaken for a clean test result.
Leave a business decision, not just a platform screenshot
Bring together attributed sales, reconciled order status, contribution, stock constraints and the proposed spend cap. If order-level margin is unavailable, say the profitability question is unresolved; do not substitute a tracker payout or a platform revenue estimate.
A useful handoff names the reporting basis, owner of missing costs, return-review date, affected products and next authorized action. That makes it possible to distinguish a traffic problem from an offer, fulfilment or measurement problem before funding more of it.