The five common return fraud patterns.
Wardrobing: customer buys, wears, returns. Common in apparel and electronics with subjective quality. Detectable through cosmetic damage on returns and customer-level repeat-return rates.
Switch-back: customer returns a different item than they ordered (an old broken one, a counterfeit, an empty box). Detectable on receipt inspection if your return process actually inspects. Most do not.
Bracketing: customer orders 3 sizes, returns 2. This is technically not fraud but operationally expensive. Detectable through order patterns where multiple variants of the same SKU are ordered and only one keeps.
Empty box: shipped with item, returned without. Detectable through weight checks at receiving (most warehouses do not weigh returns). Common with high-value small items: jewelry, electronics, perfume.
Friendly fraud: chargeback after delivery. Discussed in the chargebacks article. Pattern detection overlaps with return fraud signals.
Customer-level metrics.
A customer’s return rate is the single most informative metric. A typical Shopify customer returns 5-10% of purchases. A customer with 40% return rate is statistically anomalous; with 70% return rate is almost certainly committing some form of fraud or wardrobing.
Track per-customer: total orders, total returns, return rate, average days-to-return, share of high-value SKUs in returns. Anomalies surface fast. A customer with 4 orders in 6 months, all returned within 5 days, on a high-margin SKU, is the textbook wardrober. Action: flag for inspection on the next order or restrict to no-return-allowed checkout.
- Per-customer return rate: flag above 25%
- Days-to-return: flag clusters around the deadline (last-minute returns)
- High-value SKU concentration: flag if returns are weighted to luxury items
- Return frequency: flag customers returning monthly or more
Order-level signals.
Bracketed orders are detectable: same SKU, 3 variants (S, M, L), one customer, one cart. Many platforms allow bracketing as a deliberate strategy (Stitch Fix model). For others, bracketing is operationally expensive even if not fraud.
Address mismatch: shipping address differs from billing address by city, state, or country. Higher fraud rate. Cross-border DTC orders often see this where the billing is in one country and shipping is the actual destination. Some are legitimate (a parent buying for a child); some are stolen card.
SKU-level patterns.
Some SKUs attract more fraud than others. High-value, easily resellable items (electronics, branded apparel, jewelry) have higher return-fraud rates. Track per-SKU: return rate, average days-to-return, restock rate (how many returns make it back to sellable inventory).
A SKU with 40% return rate where 80% of returns come back damaged is almost certainly being wardrobed. A SKU with 20% return rate where 95% come back as empty boxes has switch-back fraud. The signal differs by category. Apparel: return rate 15-25% normal. Electronics: return rate 5-10% normal. Anything significantly above category baseline deserves investigation.
Return inspection workflow.
The cheapest fraud-detection mechanism is actually inspecting returns. Most warehouses do not. Implement: weigh every return on receipt against the original ship weight, photograph the return contents, check serial numbers on electronics, check tags on apparel. The marginal cost is minutes per return; the prevented loss is thousands per year.
For high-value SKUs, mandate serial-number tracking. The return must show the same serial as the original ship. Mismatch is fraud. For apparel, mandate tag-on-return policy. Cosmetic damage exceeds threshold, return rejected and refund denied (subject to local consumer law).
The accounting angle.
Returns hit gross margin twice: revenue reverses and inventory restocks. Restocked-but-unsellable units are the silent killer. If a return arrives, gets restocked at original WAC, then later gets written off because it was never sellable, the loss is delayed but not avoided.
Cleaner: at return inspection, decide sellable or not. Sellable: restock at original WAC. Unsellable: debit Inventory write-off, credit Inventory at original WAC. The write-off line surfaces the actual fraud cost. A merchant whose Inventory write-off is 1.5% of cost of goods sold has a real fraud problem; one at 0.2% probably does not.
COD-market patterns.
COD return fraud is an additional category in markets where cash-on-delivery is common. The customer accepts COD delivery, pays cash, then claims the package was empty or damaged and demands refund. The courier holds cash; reversing it requires evidence. Common with small-value, high-volume SKUs from social-commerce sellers.
Mitigation: video-record courier delivery for high-value orders, require signature on delivery, use couriers that timestamp delivery proof. Most major 3PL operators offer this. The cost is small; the deterrent is large. Customers who know delivery is recorded rarely attempt empty-box claims.
Where Nonari fits.
Nonari surfaces customer-level return rates, SKU-level return patterns, and Inventory write-off as a percentage of COGS. Anomalies appear in the dashboard without setting up custom reports. Per-customer flags can be set in checkout to require non-returnable terms or block future orders. Inventory write-offs by reason code show fraud impact separated from operational damage.
For Shopify merchants combining COD and prepaid orders, the system tracks RTO patterns alongside return patterns. A specific region showing high RTO + high empty-box claims is a fraud signal you can act on. The accounting layer surfaces what operations needs to defend margin without a separate analytics setup.