Anomaly detection & loss prevention

AI Tools → Anomaly Detection scans transactions for patterns that sit outside the norm and surfaces them for a human to look at.

What it looks for

Scoring is statistical — how far a behaviour sits from the baseline for that store and that role. Common flags:

• Voids or no-sales well above the usual rate.
• Refunds without a matching original sale, or clustered near shift end.
• Discounting far outside what colleagues apply.
• Unusual cash variance on a particular register or person.
• Transactions at odd hours for that location.

Cashier risk scores

Individual scores roll those signals into one number per staff member. Read them as where to look, never as a finding. Someone who works every closing shift will handle more refunds and more cash than a weekday colleague, and the score does not know that.

Investigating a flag

Open the flagged transaction and work outwards: the activity log shows who did what and when; the sale shows what was rung and how it was paid; the shift shows the drawer position around it. Most flags resolve into an explanation within a couple of minutes.

Managing alerts

Alerts can be reviewed and cleared so the list stays meaningful. A dashboard nobody clears stops being read, and then it may as well not exist.

Prevention beats detection

The controls that reduce loss are mostly boring: give each person their own login and PIN, restrict voids and refunds by role, require a manager for over-threshold discounts, count the drawer every shift, and act on variance while people still remember the day.

Being fair about it

These are signals, not accusations. Look into them quietly, consider the innocent explanation first, and keep in mind that most variance is process or training rather than dishonesty.

Note: your business can rename menu items (Settings → POS Configuration → Menu Labels), so the names in your menu may differ from those shown here.