Fraud & returns risk scoring

Fraud & Returns Risk Scoring

Retailos automatically scores each customer for return and payment fraud risk so you can take action before losses pile up.

How it works

The risk engine lives at /app/risk. For every customer in your store, it builds a score based on three behavioural signals:

• Serial returns — customers who return orders at an unusually high rate
• Wardrobing patterns — customers who appear to buy, use, and return items repeatedly
• COD refusal history — customers who have refused cash-on-delivery orders at the door

Alongside each score, Retailos surfaces a recommended policy action — for example, restricting a customer from the COD payment option, or requiring them to upload photos when submitting a return request.

Scores are not static. They update automatically as new behaviour accrues, so a customer's risk level can go up or down over time.

How to set it up

1. Go to /app/risk in your Retailos dashboard.
2. Review the customer list. Customers are displayed with their current risk score.
3. Click on any customer to see the specific signals driving their score and the recommended policy action.
4. Apply the recommended action — or choose a different response — directly from that customer's risk detail view.
5. Revisit /app/risk regularly. Because scores update as behaviour changes, customers you cleared previously may re-enter the list, and previously flagged customers may improve.

Good to know

• Scores are behaviour-driven, not manual. You cannot set or override a score directly; the AI determines it from actual transaction and return history.
• Recommended actions are suggestions. Retailos tells you what to consider, but you decide whether and how to act on each flag.
• Data comes from Retailos activity. If your store also runs on Shopify, order and returns data syncs two-ways between Retailos and Shopify — but the risk engine reads only the data Retailos holds.
• New customers have no score yet. A meaningful score builds up only after a customer has enough recorded interactions, so very new customers may not appear in the risk list.

Troubleshooting

A high-volume customer isn't appearing in the list. The risk engine needs sufficient transaction history to generate a score. If the customer is relatively new to your store, their score may not have been calculated yet. Check back after they have completed more orders.

The recommended action doesn't match your policy. Recommended actions are starting points based on common retail risk patterns. You are not required to follow them — apply whichever response fits your store's policies.

A customer's score hasn't updated after recent behaviour. Scores refresh as behaviour accrues, but there may be a short delay after a new return or refused COD before the score reflects it. If the issue persists, contact Retailos support.