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

# Categories

> High-level conversation classification

Categories provide high-level organization for conversation types, appearing in breakdown charts and analytics.

{/* Screenshot: Categories page showing list of categories with descriptions */}

## What are Categories?

Categories are mutually exclusive labels for conversation types. Each conversation is assigned to exactly one category.

**Example categories:**

| Category          | Description                                               |
| ----------------- | --------------------------------------------------------- |
| Billing           | Payment issues, invoices, refunds, subscription questions |
| Technical Support | Bugs, errors, how-to questions, troubleshooting           |
| Account Issues    | Login problems, password resets, profile updates          |
| Feature Requests  | Product suggestions, enhancement requests                 |
| General Inquiry   | Other questions that don't fit above                      |

## Creating Categories

<Steps>
  <Step title="Navigate to Categories">
    Go to **Tag → Categories** in your dashboard.
  </Step>

  <Step title="Click Create Category">
    Click **Create Category**.

    {/* Screenshot: Create category dialog with name and description fields */}
  </Step>

  <Step title="Enter Details">
    * **Name:** Clear, concise label
    * **Description:** What types of conversations belong here
  </Step>

  <Step title="Add Detection Hints (optional)">
    Help the AI identify this category:

    * Keywords that suggest this category
    * Example phrases or topics
  </Step>

  <Step title="Save">
    Click **Save** to create the category.
  </Step>
</Steps>

## How Auto-Classification Works

The AI assigns categories based on:

1. **Conversation content** — What the customer is asking about
2. **Detection hints** — Keywords and phrases you've provided
3. **Historical patterns** — How similar conversations were categorized
4. **Context** — The overall topic and resolution

## Designing Good Categories

### Be Mutually Exclusive

Categories should not overlap. A conversation should clearly belong to one category.

**Good:**

```
Billing (payments, invoices)
Technical (bugs, errors, how-to)
Account (login, settings)
```

**Overlapping (avoid):**

```
Payment Issues
Subscription Questions ← Overlaps with Payment Issues
```

### Be Comprehensive

Cover all conversation types. Include a catch-all for edge cases.

```
Billing
Technical Support
Account Issues
Feature Requests
General Inquiry ← Catch-all
```

### Keep It Simple

5-10 categories is usually ideal. Too many categories:

* Makes classification less accurate
* Complicates analysis
* Harder to maintain

## Using Categories in Analytics

### Breakdown Charts

{/* Screenshot: Pie chart showing category distribution */}

See what percentage of conversations fall into each category.

### Performance by Category

Compare metrics across categories:

| Category  | Volume | Resolution Rate | Avg Handle Time |
| --------- | ------ | --------------- | --------------- |
| Billing   | 35%    | 82%             | 4.2 min         |
| Technical | 40%    | 68%             | 8.5 min         |
| Account   | 15%    | 91%             | 2.1 min         |

### Drill Down

Click a category to see all conversations in that category:

{/* Screenshot: Run list filtered by category */}

## Assigning Categories to Agents

Control which categories an agent can assign:

<Steps>
  <Step title="Open Agent Configuration">
    Go to **Build → Agents** and click on an agent.
  </Step>

  <Step title="Go to Classification Tab">
    Select the **Classification** tab.
  </Step>

  <Step title="Select Categories">
    Check the categories this agent should assign.
  </Step>

  <Step title="Save">
    Save the agent configuration.
  </Step>
</Steps>

## Best Practices

* **Start with common categories** from your existing support data
* **Review periodically** — add or merge categories as needed
* **Use the catch-all** — "General Inquiry" prevents unclassified conversations
* **Include descriptions** — help the AI classify accurately
* **Check analytics** — if a category is rarely used, consider removing it

## Next Steps

<CardGroup cols={2}>
  <Card title="Attributes" icon="tags" href="/tagging/attributes">
    Add granular custom fields
  </Card>

  <Card title="Analytics Breakdown" icon="chart-pie" href="/analytics/breakdown">
    Analyze by category
  </Card>
</CardGroup>