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 gives you the raw material for every signal above. The Sales Returns report lists each credit note with its customer, the invoice it came from, its reason and the profit it took back, grouped by customer, reason or branch; Sales by Customer & Item shows what each customer buys, so a return rate per customer or per SKU is the first divided by the second. Stock write-offs carry a reason — damage, theft, expiry — so fraud losses sit apart from ordinary damage.
Blocking a customer at checkout belongs in Shopify, where the order is taken. RTO and empty-box claims come from your couriers’ reports, so keep them beside the returns data when you review a region: high RTO plus high empty-box claims in one area is a fraud signal you can act on.