RTO risk scoring
How the 0–100 score is produced, and what auto-flag and auto-cancel do.
Every COD order is scored 0–100 on address quality, order history and behavioural signals, then classified into one of four bands.
| Band | Meaning |
|---|---|
LOW | Nothing unusual. Ship it. |
MEDIUM | Worth a glance during a review pass. |
HIGH | Review before dispatch. |
CRITICAL | Strong signals of non-delivery. |
Thresholds
| Setting | Default | Effect |
|---|---|---|
rtoAutoFlagScore | 80 | Orders scoring at or above this are flagged for review. |
rtoAutoCancelScore | null | Orders scoring at or above this are cancelled before dispatch. Disabled by default. |
Auto-cancel is off by default deliberately. Start with flagging, review what lands in the queue for a few weeks, and only then decide whether a cancel threshold is warranted — and set it well above your flag score.
Audit trail
The score, its band, the full signal breakdown and the action taken are stored against every order in RtoRiskScore. A flag is never a black box: you can see exactly which signals fired, months later, when a customer asks.
Using scores well
Do not treat the score as a verdict. It is a queue-ordering tool. A HIGH score on a repeat customer with three delivered orders means something different than the same score on a first-time buyer.
Review the flagged queue on a schedule. A flag that nobody looks at is just a delayed dispatch. Ten minutes daily beats an hour weekly, because the value of the flag decays with the shipping window.
Recheck your thresholds each quarter. As your customer mix changes, the score distribution moves with it.