Every retailer measures the sale. Almost nobody measures the part that comes after it. A customer buys, the pixel fires, ROAS ticks up, and three weeks later a third of that order is back in the warehouse. The ad platform never sees the return. It keeps spending against a number that already happened and won't hold.
Returns are a profitability drain that sits entirely outside the measurement window. They land after the point where everything gets counted, and they vary more from product to product than almost any other signal you optimize on.
The ROAS That Doesn't Survive the Return
Picture two products that look identical on the dashboard. Both sell at €100, both report a 3.0 ROAS in Google Ads. One is a dress with a 38% return rate. The other is a low-return item, say a kitchen gadget, that comes back 6% of the time.
The dashboard treats them as equal. The bank account doesn't.
On the dress, only 62% of the revenue actually stays. After returns, the 3.0 headline number is closer to 1.9. And that is before the cost of processing the return, restocking it, and writing off what comes back unsellable. On the kitchen gadget, 94% of the revenue sticks, so the same 3.0 is closer to 2.8. Same dashboard number; one product makes roughly half the profit of the other per euro of ad spend.
PMax can't see any of that. It optimizes on the 3.0 it can measure, so it keeps funding the dress at least as hard as the gadget, and often harder if the dress's surface ROAS runs a touch higher. The budget flows toward the product that comes back, because returns are invisible at the exact moment the algorithm makes its decision.
This is the same gap that makes ROAS misleading on margin, one step further out. ROAS doesn't see your cost of goods, and it doesn't see your returns either. The difference is that returns can swing the real number by half, and they do it on precisely the categories that already carry the thinnest post-return economics.
Returns Vary More Than the Signal You're Optimizing On
The reason this matters so much in retail is the spread. Return rates aren't a small correction applied evenly across the catalog. They range from rounding error to nearly half of revenue depending on what you sell.
| Category | Typical online return rate |
|---|---|
| Apparel & fashion | 25–40% |
| Footwear | 20–30% |
| Accessories | 10–15% |
| Electronics | 8–15% |
| Home & garden | 10–20% |
| Books & media | 3–5% |
Overall online returns run around 20% (NRF's 2025 Retail Returns Landscape), and the spread across categories is wide. Apparel and footwear sit at the top because sizing drives most of their returns and 'bracketing', ordering several sizes to keep one and send the rest back, is now mainstream. Standardized categories like electronics and books run lower. These are rough ranges and yours will differ, but the shape holds everywhere: a multi-category retailer is blending economics that have nothing to do with each other. If 40% of your catalog is apparel and 10% is books, your 'average' return rate is a number that describes no actual product.
That blended average is what hides the problem. A dress with a 38% return rate can be losing money on every ad-acquired sale while a low-return product two rows down in the same report is genuinely profitable. Looked at together, they average out to something that looks fine, and the budget keeps flowing on the surface ROAS.
Why the Platforms Can't See It
Returns happen after the purchase. When a customer sends an item back, that event lives in your ecommerce system, your warehouse, your payment processor, and your returns software. Unless you explicitly send that information back, the ad platform saw the purchase pixel fire weeks earlier and counted the revenue. Nothing tells it that a chunk of the sale reversed.
Google Ads does support conversion adjustments for returned purchases and partial returns, and those adjustments can affect CPA and ROAS bid strategies. That is useful for correcting reported conversion value. It is still not the same as a product-level merchandising rule. It does not say: products with high returns and thin margins should receive lower priority before the next order happens.
So the product-level return rate sits in your data, fully known to you and usually invisible to the system spending your budget.
Return Rate as a Priority Signal
The fix is not to create another dashboard metric and hope the platform understands the business context behind it. The fix is to use return rate as one of the signals that decides where budget goes, before it ever reaches the platform.
Return rate is a field your Business Rules can read, sitting next to margin, stock cover, and ROAS. You connect your return data as a feed, at product or variant level, or as category-level values mapped back to products. From there, the logic is the kind of thing any merchandiser would recognize:
- High return rate and thin margin → Low priority. The product is a profit leak; stop over-funding it.
- High return rate but high margin or strategic importance → Standard. Manage it, don't kill it.
- Low return rate and healthy margin → High priority. This is the revenue that actually stays.
Return rate behaves like any other number in your Business Rules. Set a hard threshold: anything above 30% drops a tier. Or set a relative rule: the worst-returning tenth of the catalog gets deprioritized. Combine either with margin or stock cover. The rules re-evaluate on each scheduled run. Fresh return data updates the value. When a product crosses your threshold, it gets pulled back on the next sync. Nobody touches the campaign.
The mechanism matters, because this is easy to overclaim. Expanly does not rewrite your ROAS or compute a single 'true' efficiency number for Google. It uses the return signal to set a priority label. That label flows to your Google Merchant Center feed. Your campaign structure can use it to route spend. Your return numbers stay inside the rule logic; only the priority label reaches Google. The dashboard metric does not change. The money moves. Over time, the blended return rate on ad-driven sales falls because the budget stops chasing the products that come back.
The return signal also rarely works alone. The same product carries a margin, a stock position, and a place in your seasonal plan. A high-return product that's also high-margin and central to the season is a different decision from a high-return product that's thin-margin and replaceable. Return rate earns a place in the priority decision; it doesn't override everything else.
What It Looks Like in Practice
One omnichannel fashion account showed the pattern clearly: a broad assortment, stores and ecommerce, and product-level returns that changed the economics after the ad platform had already counted the sale. Their reporting showed surface ROAS, but not which ad-driven products were most likely to come back.
When return rate became part of the priority logic, budget started shifting away from products with persistently high returns and weak economics. Over comparable year-over-year periods, their blended e-commerce return rate was about 10 percentage points lower. The exact shape moved month to month, as it always does in fashion, but the commercial effect was clear: fewer ad euros went to the products most likely to be returned.
Two things improved at once. The obvious one is fewer returns, which means less reverse logistics, less restocking, and fewer items coming back unsellable. The quieter one is customer experience: a return is a small failure, and steering demand toward products that fit and stick means more first orders that go right. Lower returns improve profitability twice over, once on the cost side and once on the repeat-purchase side.
