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

# Set proactive alerts for support quality

> Configure threshold-based and content-based alerts that notify you when CSAT drops, bad ratings spike, AI confidence falls, or ticket volumes surge.

Proactive alerts let you catch support quality problems before they escalate. Instead of discovering a CSAT drop in your weekly review, raum notifies you the moment a metric crosses a threshold you've defined — or when specific patterns appear in ticket content. You set the rules; raum watches continuously.

## What alerts monitor

raum supports two categories of alerts:

**Metric-based alert rules** watch your aggregated support metrics and fire when a threshold is crossed:

* CSAT score drops below a defined level
* Bad rating rate exceeds a defined percentage
* AI confidence score falls below a threshold (a leading indicator that your procedures need updating)
* Ticket volume spikes above a defined count within a time window

**Content-based alerts** scan ticket content using pattern matching. You define the keywords or phrases to watch for, and raum notifies you whenever those patterns appear. This is useful for tracking competitor mentions, specific product issues, legal language, or any topic you want early visibility on.

## Alert rule types

<CardGroup cols={2}>
  <Card title="CSAT drop" icon="face-frown">
    Fires when your average CSAT score falls below a threshold you set. Configure the threshold as a percentage and the measurement window (for example, last 24 hours or last 7 days).
  </Card>

  <Card title="Bad rating rate" icon="thumbs-down">
    Fires when the proportion of negative ratings exceeds a defined percentage. Useful for catching a sudden influx of dissatisfied customers before your overall CSAT is significantly affected.
  </Card>

  <Card title="Low AI confidence" icon="brain">
    Fires when the AI's average confidence score drops below a threshold. This typically means new ticket types are arriving that your procedures don't cover well — a leading indicator before CSAT suffers.
  </Card>

  <Card title="Volume spike" icon="chart-line">
    Fires when ticket volume exceeds a count within a time window. Use this to get ahead of incident-driven surges so you can activate additional agents or prepare a temporary update.
  </Card>

  <Card title="Content pattern" icon="magnifying-glass">
    Matches keywords or phrases in incoming ticket content. Fires whenever a matching ticket arrives. Configure multiple patterns per alert rule.
  </Card>
</CardGroup>

## Creating an alert

Alert rules are configured at `/dashboard/alerts-content`.

<Steps>
  <Step title="Open alerts configuration">
    Navigate to `/dashboard/alerts-content` in your dashboard. You'll see your existing alert rules and a button to add a new one.
  </Step>

  <Step title="Choose the alert type">
    Select the metric or content type you want to monitor. For metric-based alerts (CSAT, bad rating rate, confidence, volume), you'll set a threshold and a measurement window. For content-based alerts, you'll enter the patterns to match.
  </Step>

  <Step title="Set the threshold or patterns">
    For metric alerts, enter the threshold value. For example, to alert when CSAT drops below 80%, set the threshold to 80. For content alerts, enter the keywords or phrases — one per line. Matching is case-insensitive.
  </Step>

  <Step title="Configure the cooldown">
    Set a cooldown period to prevent alert spam. The cooldown is the minimum time between two alerts of the same rule. For example, a 4-hour cooldown means you'll receive at most one notification per rule every 4 hours, even if the threshold remains exceeded.
  </Step>

  <Step title="Save and activate">
    Save the rule. It activates immediately and begins monitoring. You can pause or delete rules at any time from the alerts configuration page.
  </Step>
</Steps>

## Alert cooldowns

Without cooldowns, a sustained threshold violation would generate a continuous stream of notifications. The cooldown period ensures each alert rule fires at most once per interval, regardless of how many threshold-crossing events occur.

<AccordionGroup>
  <Accordion title="How cooldowns work">
    When an alert fires, raum records the timestamp. The rule is suppressed until the cooldown period elapses. After the cooldown, if the condition is still met, the alert fires again. This means you stay informed without being flooded.
  </Accordion>

  <Accordion title="Choosing the right cooldown">
    For CSAT and confidence alerts, a 4–8 hour cooldown is usually appropriate — long enough to avoid noise, short enough to re-alert if a problem persists into the next shift. For volume spikes caused by incidents, a shorter cooldown (30–60 minutes) may make sense so you get reminded if the spike continues.
  </Accordion>

  <Accordion title="Content alerts and cooldowns">
    Content-based alerts can fire very frequently if a common phrase is matched. Set a meaningful cooldown unless you genuinely want per-ticket notification. For critical patterns (legal language, urgent safety issues), a short cooldown or no cooldown may be appropriate.
  </Accordion>
</AccordionGroup>

## How alerts are delivered

When an alert fires, raum records the event and delivers a notification. The notification includes:

* The alert rule name
* The metric value or matched content that triggered it
* A timestamp
* A link to the relevant dashboard section

<Note>
  Notification delivery channels (email, Slack, webhook) depend on your plan and notification settings. Configure your preferred channels in `/dashboard/settings`.
</Note>