Return Fraud Prevention
What actually works in 2026 — and what doesn't. A practical breakdown for SMB and mid-market merchants.
The scale of the problem
The National Retail Federation estimates return fraud cost US retailers around $101B in 2023 — roughly 13.7% of all returns. For online merchants, the numbers are higher: e-commerce return rates run 20-30% with fraud accounting for 10-15% of those.
For a mid-market merchant doing $5M/year, that's typically $50-100k/year in absorbed losses that show up as "support cost" or "cost of doing business" rather than fraud.
The 6 patterns that drive most losses
Wardrobing
Buying an item, using it briefly (a dress for one event, a camera for one trip), and returning it as "unwanted." Common in apparel, electronics, and luxury goods.
Bracketing
Ordering 5 sizes or 3 colors, keeping the one that fits, returning the rest. Legitimate on paper, abusive in practice — costs merchants the shipping, restocking, and inventory churn on 4 of 5 items.
Serial refund abuse
Requesting refunds on delivered items by claiming the product was defective, damaged, or never received. The product is kept. Refund is granted. Cycle repeats across merchants.
Receipt and policy abuse
Returning items bought elsewhere, using receipts from past purchases, or exploiting return windows that overlap with sale cycles to pocket the difference.
Empty box returns
Returning a box with the packaging and no product, or a substitute item. Works when warehouses don't inspect returns carefully. More common in high-volume retailers than small merchants.
Chargeback-after-refund
Getting a refund approved, then filing a chargeback claiming the charge was unauthorized. Merchant ends up out the product AND the money. Rarer but expensive when it hits.
Why strict return policies fail
The instinct to tighten return policies in response to fraud is understandable and mostly wrong. Strict policies punish the 85% of customers who are legitimate and lightly inconvenience the 15% who are abusing the system.
Serial refund abusers specifically seek out merchants with generous return policies. When you tighten, you lose legitimate revenue from customers who fear they can't return something that genuinely doesn't fit. You don't deter abusers — they just move to the next merchant.
The data backs this up: merchants who tightened return windows in 2022-2023 saw modest fraud drops and larger cart abandonment increases. Net revenue impact was negative for most.
What actually works
- Score refunds, don't block them. Every refund request gets a risk score based on card history, device/IP patterns, authentication data, and timing. Most refunds still go through instantly. High-risk ones get friction (photo request, explanation) before approval.
- Use cross-merchant signal. Single-merchant data misses serial abusers who spread their activity across 10 merchants. Networks that share hashed signals (device, IP, card fingerprint) surface patterns individual merchants can't see.
- Track authentication data at purchase. 3DS, AVS, and CVC results captured at the original charge are strong evidence against later refund claims. If the customer authenticated their own card, the "I didn't make this purchase" narrative is weaker.
- Require proof for high-risk refunds only. Asking every customer for a photo burns goodwill. Asking only customers whose refund request scores high (based on signals they don't know you have) is efficient and surprisingly effective — most serial abusers ghost when asked for proof.
How CertNode Sentinel handles this
Sentinel scores every refund request for Stripe merchants using the 5 signals above, plus a cross-merchant abuse network that aggregates hashed signals from every Sentinel merchant's refund outcomes.
You see the score in your Stripe Dashboard when you open the charge. Low-risk refunds show Approve. Medium-risk show Request Proof. High-risk show Pause. You always make the final call.
$49/month flat. $39/month with the Reflex bundle. 14 days free. Install from the Stripe App Marketplace.