Most store dashboards show returns as one percentage: units returned divided by units sold. That number hides almost everything that matters. A 25 percent return rate on a product that restocks cleanly can cost less than a 10 percent rate on a product that comes back damaged. A low return rate this month can simply mean this month’s orders have not had time to come back yet.
The scale makes this worth getting right. The National Retail Federation and Happy Returns estimate that returns will total $849.9 billion in 2025, with 19.3 percent of online sales sent back. Across all channels, retailers expect 15.8 percent of annual sales to be returned. Processing is expensive too: Optoro, a reverse logistics company, puts the average cost of a return at 27 percent of the purchase price, enough to erase as much as half of the sales margin.
Where the money leaks
A refund reverses revenue, but the cost of a return builds up across several steps, and each step leaks a different amount depending on the product.
Seven metrics instead of one
| Metric | Definition | What it tells you |
|---|---|---|
| Unit return rate | Units returned divided by units shipped, for a closed cohort of orders | How often the product disappoints |
| Value return rate | Refunded merchandise value divided by merchandise revenue | Whether expensive items come back more often than cheap ones |
| Refund without return | Refunds with no item received (goodwill, damage claims, “keep it”) | Policy cost, kept separate from product quality |
| Recovery rate | Value recovered from returned units divided by their cost of goods | How much of a return you get back |
| Cost per return | Labels, labour, inspection, packaging, divided by returns processed | The handling bill, usually larger than expected |
| Exchange share | Returns converted to exchanges or store credit | Revenue kept despite the return |
| Return-adjusted contribution | Contribution per unit shipped after all of the above | The number to rank products by |
Return-adjusted contribution, worked through
For each unit shipped, a kept unit earns price minus cost of goods. A returned unit earns nothing, loses the share of its cost of goods that cannot be recovered, and adds handling cost. With price P, cost of goods C, unit return rate r, recovery rate k and handling cost per return h:
Contribution per unit shipped = (1 - r) x (P - C) - r x ((1 - k) x C + h)
Two hypothetical products, both selling at $60, with handling cost set at Optoro’s 27 percent average ($16.20):
| Product A (apparel) | Product B (accessory) | |
|---|---|---|
| Price / cost of goods | $60 / $24 | $60 / $30 |
| Gross margin per unit | $36.00 | $30.00 |
| Unit return rate | 25% | 5% |
| Recovery rate | 80% | 90% |
| Margin from kept units | 0.75 x $36 = $27.00 | 0.95 x $30 = $28.50 |
| Return cost per unit shipped | 0.25 x ($4.80 + $16.20) = $5.25 | 0.05 x ($3.00 + $16.20) = $0.96 |
| Return-adjusted contribution | $21.75 | $27.54 |
Product A wins on gross margin and loses by almost $6 a unit once returns are counted. A merchandising report sorted by gross margin would promote the wrong product.
Measure returns on closed cohorts
Returns arrive weeks after the sale. If you divide this month’s returns by this month’s sales, the recent months always look better than they are, and a holiday peak distorts the following quarter. Group orders by the month they shipped and only report a cohort’s return rate once its return window has closed (for example 45 days for a 30-day policy, to allow for transit). This is the same logic as a retention table, with “returned” in place of “came back to buy”. QuantumLayers walks through building and reading that kind of table in Cohort Analysis: How to Build a Retention Table.
Getting the data out of WooCommerce and Magento
Both platforms record refunds well and returns poorly. In WooCommerce, each refund is its own record of type shop_order_refund linked to the original order, as described in the High-Performance Order Storage schema. In Magento, refunds are credit memos (sales_creditmemo and sales_creditmemo_item). Neither tells you on its own whether an item physically came back, what condition it was in, or why.
Three small process changes fix most of that:
- Reason codes. Replace free text with a short fixed list (wrong size, not as described, damaged in transit, defective, changed mind, ordered several to compare). Free text cannot be grouped.
- Disposition. Record what happened to each returned unit: restocked, open box, liquidated, written off. This is what gives you a recovery rate.
- Refund type. Flag refunds with no item returned so they do not inflate the product’s return rate.
Segments that change decisions
- By product and variant. A size that is returned far more than its neighbours is usually a sizing chart problem, not a customer problem.
- By reason. “Not as described” points at photos and copy. “Damaged in transit” points at packaging and carriers.
- Multi-size orders. Orders containing the same product in two or more sizes are the footprint of bracketing, the habit of buying several options and returning the rest, which CNBC reports is adding to retailers’ cost pressure. Measure its share before deciding whether to discourage it.
- By customer. Return behaviour is concentrated. NRF found shoppers aged 18 to 30 made an average of 7.7 online returns in the past year, more than any other age group. Plot the distribution of customer-level return rates and look at the top few percent before changing policy for everyone.
- By acquisition channel. A channel with a high conversion rate and a high return rate may be worth less than its revenue suggests.
Fraud belongs on the list too, kept as its own line: the same NRF report estimates that 9 percent of all returns are fraudulent. Merging fraud into the general return rate makes both harder to act on.
A monthly returns review in four questions
- Which products have the lowest return-adjusted contribution, and is it the return rate or the recovery rate that pulls them down?
- Which reason codes grew in the latest closed cohort?
- What share of refunds had no item returned, and is that share rising?
- Has the customer-level distribution changed shape, or only its average?
None of these questions can be answered from a single return-rate figure, and all of them can be answered from data most stores already have once reason codes and dispositions are recorded.
Sources
- National Retail Federation and Happy Returns, 2025 Retail Returns Landscape
- The Dead Pixels Society, NRF: Consumers expected to return nearly $850 billion in merchandise in 2025
- AJOT, Consumers expected to return nearly $850 billion in merchandise in 2025
- Optoro, Optoro launches Returns Unwrapped
- CNBC via NBC Los Angeles, Retailers are rushing returns back to resale market as tariff costs hit item economics
- WooCommerce Developer Blog, High-Performance Order Storage: Database Schema


