> ## Documentation Index
> Fetch the complete documentation index at: https://docs.flashcat.cloud/llms.txt
> Use this file to discover all available pages before exploring further.

# Outlier Incidents

> Outlier incident detection for quick handling of new incidents

<Tip>**Plan requirement**: This feature requires an On-call Pro or higher subscription. [Learn more](https://flashcat.cloud/flashduty/price/)</Tip>

An outlier incident is one that has never occurred in the past period. If an incident is identified as an outlier, responders should be alert. This is critical for On-call engineers because outlier incidents may have unknown impacts, and handling procedures may need to be decided on the spot. Some engineers may only care about outlier incidents—they create standardized processes or SOPs for resolving such incidents.

## View Outlier Incidents

<Tabs>
  <Tab title="Console">
    1. On the incident list page, outlier incidents display a prominent **Outlier** badge
    2. On incident details page, outlier incidents show a prominent **Outlier** badge and explanation at the top

    <Frame>
      <img src="https://download.flashcat.cloud/flashduty/doc/en/fd/outlier-1.png" alt="Console Outlier Incident Badge" />
    </Frame>
  </Tab>

  <Tab title="IM Tools">
    Using Feishu/Lark as an example, when an incident is identified as an outlier alert, the system enhances the reminder:

    <Frame>
      <img src="https://download.flashcat.cloud/flashduty/doc/en/fd/outlier-2.png" alt="Feishu/Lark Outlier Incident Reminder" />
    </Frame>
  </Tab>
</Tabs>

### Recognition Method

The system uses machine learning models to determine similarity between incidents. When similarity exceeds 90%, we consider two incidents similar.

When judging similarity, we mainly consider these factors:

| Factor                      | Description                                |
| :-------------------------- | :----------------------------------------- |
| Incident Title              | Semantic similarity of title text          |
| Incident Description        | Semantic similarity of description content |
| Affected Service            | Generally extracted from `service` label   |
| Alert Resources in Incident | Generally extracted from `resource` label  |

<Note>
  When the system detects no similar incidents in the past **30** days, it marks the incident as an outlier.
</Note>

## Enable and Disable

Outlier incident detection is configured at the channel level:

<Steps>
  <Step title="Enter Channel Settings">
    Go to **Channel** → Select target channel → **Basic Settings**
  </Step>

  <Step title="Configure Detection Toggle">
    Find **Outlier Incident Detection** option, enable or disable the feature
  </Step>
</Steps>

<Tip>
  When enabled, incident lists and notification content will include an "Outlier" badge for quick identification. See [Create and Manage Channels](/en/on-call/channel/create-edit) for details.
</Tip>

## FAQ

<AccordionGroup>
  <Accordion title="Does outlier incident detection need to be enabled separately?">
    Enabled by default for Pro and above plans, but can be disabled in the channel's basic settings. See the "Enable and disable" section above.
  </Accordion>
</AccordionGroup>

## Related Topics

<CardGroup cols={2}>
  <Card title="Past Incidents" icon="clock-rotate-left" href="/en/on-call/incident/past-incidents">
    Review historical similar incident resolutions
  </Card>

  <Card title="What is an Incident" icon="circle-exclamation" href="/en/on-call/incident/what-is-incident">
    Understand incident basics and lifecycle
  </Card>
</CardGroup>