---
title: "Set evaluation behavior and labels | Grafana Labs"
description: "Configure how often your alert is evaluated and organize it with labels."
---

# Set evaluation behavior and labels

The evaluation behavior determines how often Grafana checks your alert condition and how long the condition must persist before firing. Labels help organize your alerts and drive notification routing.

To configure evaluation behavior and labels, complete the following steps:

1. In the **Set evaluation behavior** section, click **Select folder** to choose or create a folder for your alert rule.
   
   Folders help organize alert rules by team, service, or purpose.
   
   For example, select or create a folder named `Infrastructure alerts` or `Application logs`.
2. Select or create an **Evaluation group**.
   
   Evaluation groups batch rules that share the same evaluation interval.
   
   Expand table
   
   | If alerting on | Example group name      |
   |----------------|-------------------------|
   | Metrics        | `CPU and memory alerts` |
   | Logs           | `Error rate alerts`     |
3. Set the **Evaluation interval**—how often Grafana runs your query and checks the condition.
   
   For example, enter `1m` for critical alerts or `5m` for less urgent monitoring.
4. Set the **Pending period** (also called “for” duration)—how long the condition must be true before the alert fires.
   
   Expand table
   
   | Alert type          | Pending period | Why                                               |
   |---------------------|----------------|---------------------------------------------------|
   | Metrics (CPU spike) | `5m`           | Filter brief spikes during deployments            |
   | Logs (error burst)  | `2m`           | Catch sustained error patterns, not single errors |
   
   > **Tip:** A 5-minute pending period with a 1-minute evaluation means the condition must be true for 5 consecutive checks before firing.
5. In the **Add labels** section, click **Add labels** to add labels that identify and categorize your alert.
   
   - Click **+ Add label**.
   - Enter a label key and value.
     
     For example:
     
     - `severity`: `warning` or `critical`
     - `team`: `platform` or `database`
     - `type`: `metrics` or `logs`
     - `service`: `api-gateway` or `payment-service`
6. In the **Add annotations** section, optionally add context for responders:
   
   - **Summary**: Brief description of what’s wrong.
     
     Expand table
     
     | If alerting on | Example summary                               |
     |----------------|-----------------------------------------------|
     | Metrics        | `CPU usage is high on {{ $labels.instance }}` |
     | Logs           | `Error rate elevated for {{ $labels.job }}`   |
   - **Description**: Detailed context or troubleshooting steps.
   - **Runbook URL**: Link to documentation for responding to this alert.

In the next milestone, you’ll configure where notifications are sent when this alert fires.

* * *

### More to explore (optional)

- [Labels and label matchers](/docs/grafana/latest/alerting/fundamentals/alert-rules/annotation-label/)