Skip to main content

Introduction

The Anomaly Detection module watches your store and flags unusual patterns in real time. It runs every hour on a schedule and writes any findings to a dedicated log that admins can review and resolve. When an anomaly is detected, the platform automatically emails every admin who has opted in to notifications, and surfaces the same finding in the in-app notification bell.
No configuration is required — anomaly detection runs out of the box on every install. You only need to opt admins in to email notifications if you want alerts in your inbox.

What Gets Detected

Severity Levels

Each finding is automatically classified by how far it deviates from the baseline:

Enabling Email Notifications for an Admin

To receive email alerts when an anomaly is detected:
1

Open the Admin Panel

Click on the Admin Panel.
2

Open Staff / Users Management

Navigate to the staff list where admin accounts are managed.
3

Edit an Admin

Open the admin’s profile and make sure their email address is set.
4

Enable Email Notifications

Toggle Enable Email Notifications on for that admin.
5

Save Changes

Click Save.
Only admins with both an email address and the Enable Email Notifications flag receive anomaly emails. The in-app notification bell receives the same alerts regardless of the email flag.

Reviewing Anomalies

To review detected anomalies:
1

Open Statistics

Click on the Statistics tab in the sidebar.
2

Open Anomalies

Open the Anomalies sub-section.
3

Investigate Findings

Each anomaly shows its type, severity, a description comparing today’s value against the historical baseline, and the raw numbers that triggered it.
4

Resolve or Mark False Positive

Use the actions on the anomaly’s detail page to move it to Resolved or False Positive, or leave it in Investigating while you dig in.

Anomaly Statuses

Running the Detection Manually

The scheduler runs anomaly detection every hour automatically. To trigger a run manually (e.g. while testing), execute the artisan command:
Manually running the detector creates real anomaly records (and may send real emails) — it is not a dry run. Use it only when you intend to flag findings for the current state of the store.

Quiet Periods

The detectors are designed to fail quiet:
  • A revenue drop is not flagged on a fresh store with no historical baseline.
  • A product surge requires at least 5 sales to be flagged.
  • A geographic surge requires at least 3 sales from a new country (or 5 from an existing one).
  • A transaction velocity spike requires at least one transaction in the past 7 days as a baseline.
This avoids spamming admins during low-traffic launch periods.