The returns math: what wrong-fit and wrong-part returns really cost
Returns aren't just a shipping cost. Here's the full model - reverse logistics to lost trust - and how much of it traces back to bad product data.
Ask a retailer for their return rate and you'll get an answer to the decimal point. Ask what's driving it, and the room goes quiet.
That's not a data gap. The reason codes already exist inside the returns system — they're just never joined back to the product record that caused the return in the first place. Do that join, build out the full cost stack, and the returns line stops looking like a fulfillment problem. It starts looking like a product-data problem that happens to arrive with a fulfillment bill attached.
The real return rate, and why it's worse than the headline number
Ecommerce returns now sit around 19-21% of orders on average — roughly two to three times the brick-and-mortar rate of under 9% — according to the National Retail Federation's 2025 Retail Returns Landscape data as summarized by Richpanel. Category spread is wide:
- Apparel: 20-40%
- Footwear: 17-30%
- Home and furniture: 15-23%
- Electronics: 8-15%
- Beauty: 4-12%
U.S. retail returns totaled roughly $890 billion in 2025 by some estimates — close to 15% of annual sales, per the same NRF-sourced dataset. That's the headline number. The part that doesn't make the press release: a large share of those returns were preventable at the point of listing, not the point of shipping.
Building the full cost stack
A "return" isn't one line item. Every unit that comes back triggers a sequence of costs, and most finance teams only track the first one or two of them.
| Cost bucket | What it captures | How to measure it |
|---|---|---|
| Reverse logistics | Inbound shipping, warehouse handling, label costs | Freight + 3PL invoices tagged to return SKUs, per Shopify's reverse logistics breakdown |
| Inspection & restocking | QA check, repackaging, putaway labor | Labor hours per return x fully loaded hourly rate |
| Markdown / liquidation | Value lost when a returned unit can't be resold at full price | (Original price - resale/liquidation price) x units returned |
| Lost margin on the original sale | Contribution margin given up when the unit never nets out as a keep | Unit margin x return rate; compounds fast — a 25% return rate can cut unit contribution margin dramatically |
| CX and support load | Agent time on return authorizations, disputes, refund processing | Support tickets tagged "return/exchange" ÷ total return volume, cost per ticket from your helpdesk |
| Trust and repeat-purchase erosion | Customers who don't come back after a bad return experience | Repeat purchase rate, cohort-split by return-history vs. no-return-history |
Processing a single return typically costs somewhere between $10 and $65 depending on category and channel, with electronics and furniture at the high end because of size, weight, and refurb requirements. Layer in that two-thirds of shoppers say they won't buy from a retailer again after a poor return experience, and the real cost model runs well past the current quarter's P&L. That last row is the one most retailers never quantify — and it's often the largest number on the sheet.
How much of that is a product-data problem
Not every return is a data problem. Some are genuine change-of-mind, some are damage in transit, some are just buyer's remorse. But the returns industry's own reason-code data points to a data-quality core:
- Fit and sizing is consistently cited as the single largest return driver in apparel, with estimates running from roughly 60% to 70%-plus of apparel returns tied to size or fit mismatch.
- "Item different than described or pictured" shows up as a top-three reason code across general merchandise, frequently in the 20-25% range of stated return reasons.
- Wrong item received — a fulfillment-adjacent but often catalog-rooted issue (duplicate SKUs, mismatched variant mapping, ambiguous product identifiers) — is regularly cited in a similar range.
Add those up and a meaningful share of returns in any non-trivial catalog — plausibly a third to half in apparel and general merchandise — trace back to the product record. A missing dimension. An out-of-date size chart. A spec that didn't match the buyer's actual need. An image that implied a color or finish the item didn't have.
That's not a shipping problem or a customer problem. That's an enrichment problem — and it shows up on the returns dashboard, not the PDP dashboard, which is exactly why it goes unmeasured.
The attributes that move the needle most
Not all missing fields cost the same. Some attributes correlate directly with return reason codes; others are nice-to-have. Prioritize enrichment against the return driver — not against catalog completeness for its own sake.
| Attribute type | Return driver it addresses | Reported impact |
|---|---|---|
| Size chart / fit measurements | Wrong size, fit mismatch | Detailed, brand-normalized size charts are associated with return-rate reductions in the 20-30%+ range in fashion ecommerce studies |
| True material / fabric spec | "Not as described" | Reduces expectation-gap returns when material, weight, and stretch are stated plainly, not just "cotton blend" |
| Compatibility / fitment data | Wrong part, wrong SKU purchased | Critical in auto parts, appliance parts, and hardware, where fitment errors drive a large share of the ~19% category return rate |
| Accurate, multi-angle imagery + color-true rendering | "Looks different than pictured" | Directly targets the described-vs-received gap, one of the most cited return reasons |
| Dimensions and weight | Space/fit disappointment in furniture and home | Reduces the 15-23% home-goods return band tied to size-in-room mismatches |
The pattern holds across categories: attributes that close the gap between what the buyer expected and what arrived are the ones with measurable return impact. Attributes that just make a listing longer are not.
Measuring the connection at your own company
You don't need a new BI stack to see this. Three steps:
- Pull your return reason codes for the last two quarters.
- Bucket them into "data-attributable" (size/fit, not-as-described, wrong item/spec) versus "non-data" (damage, change of mind, delivery).
- Join that against a data-completeness or data-quality score for the returned SKUs.
If data-attributable returns cluster on the SKUs with the thinnest, oldest, or lowest-scored attribute sets — and they usually do — you've got a defensible ROI case. Model the reverse-logistics-plus-CX cost of just the data-attributable slice, then price out closing the attribute gap on your highest-return, highest-volume SKUs first.
This is the same problem Anglera exists to close on the front end. Your PIM stores the size chart, the fitment data, the material spec. Anglera continuously scores what's missing or stale, gap-fills it from supplier and source documentation, and keeps it current as products and variants change — without replacing whatever system already holds the data.
Treat returns math as a second, brutally honest measurement of catalog quality. PDP conversion tells you if the listing gets the sale. The returns line tells you if the listing was actually true.