Skip to main content
By combining log filters with Topline Alerts, you will know immediately when things start failing without writing custom scripts or dashboards.

​
When to Use Log-based Alerts

  • Monitoring success rates - Catch regressions before they impact SLOs.
  • Detecting error spikes - Trigger when 5xx or other errors rise above baseline.
  • Isolate by segment - Identify failures concentrated in by region, client, or device type

​
Create a Log-based Topline Alert (Statsig Cloud)

In this example, we’re going to create a monitor for the success rate of a GET request.
1

Navigate to Topline Alerts

  • Go to Analytics → Topline Alerts in the product menu.
  • Click +Create.
  • Enter a name for your alert.
Alert setup step 1
2

Define Alert Criteria

  • Select statsig::log_line as your event.
  • Apply filters to define what constitutes success and failure.
If you’re unsure which fields to filter on, open the Logs Explorer and inspect the log body.
Alert setup step 2
3

Set Up Evaluation Formula

  • Define the formula for calculating your success rate.
  • (Optional) Add a group-by dimension.
    • Alert setup step 3
    4

    Configure Alert Thresholds

  • Set thresholds and the evaluation window.

    • In this example, we will trigger a warning when success rate drops below 99.5%. We will trigger an alert when success rate drops below 99.0%.

    Alert setup step 4
    5

    Configure Notifications

    • Add the alert title and description for context.
    • Include diagnostic hints (e.g. “Check version X” or “Android requests timing out”).
    • Add subscribers.
    • Set priority.
    Alert setup step 5
    6

    Save the Alert

  • Click Save.
  • Once your alert is set up, you can visit the Diagnostics tab to see a history of alert triggers.
    • Alert setup step 6

    ​
    Best Practices

    • Set up Slack notifications for team visibility
    • Keep formulas simple (ratios & percentages are easiest to scan)
    • Add group-by dimensions (like country or app version) to pinpoint where issues occur
    • Write clear notification text that explains what the alert means