> ## 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.

# Resolution Rules

> Define what "resolved" means for your team

Resolution rules define the conditions under which a conversation is considered "resolved" — critical for tracking deflection and resolution rates.

{/* Screenshot: Resolution rules page showing configured rules */}

## What are Resolution Rules?

Resolution rules are conditions that determine when a conversation is successfully resolved:

* Customer confirmed the solution worked
* Issue was fixed and customer expressed satisfaction
* Informational question was answered
* Time elapsed without follow-up

## Why Resolution Matters

Resolution tracking powers key metrics:

| Metric                 | Calculation                          |
| ---------------------- | ------------------------------------ |
| **Resolution Rate**    | Resolved ÷ Total Conversations       |
| **Deflection Rate**    | Resolved by AI ÷ Total Conversations |
| **Time to Resolution** | First message → Resolution           |

Without resolution rules, you can't measure success.

## Creating Resolution Rules

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

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

    {/* Screenshot: Create resolution rule dialog */}
  </Step>

  <Step title="Name the Rule">
    Give it a descriptive name:

    * "Customer Confirmed"
    * "No Follow-up Needed"
    * "Time-based Resolution"
  </Step>

  <Step title="Define Conditions">
    Specify when this rule marks a conversation as resolved.
  </Step>

  <Step title="Save">
    Click **Save** to activate the rule.
  </Step>
</Steps>

## Condition Types

### AI-Based

Use natural language to describe resolution criteria:

{/* Screenshot: AI-based resolution rule configuration */}

**Example prompts:**

```
Mark as resolved when the customer explicitly confirms their 
issue is fixed or expresses satisfaction with the resolution. 
Look for phrases like:
- "That worked!"
- "Thanks, that fixed it"
- "Perfect, I'm all set"
- "Great, that's exactly what I needed"
```

```
Mark as resolved when the agent successfully completed 
the requested action and the customer acknowledged it.
```

### Time-Based

Resolve after a period of inactivity:

{/* Screenshot: Time-based resolution rule configuration */}

| Setting               | Example                          |
| --------------------- | -------------------------------- |
| **Condition**         | No customer response for X hours |
| **After agent reply** | After the agent's last message   |
| **Time period**       | 24 hours, 48 hours, 72 hours     |

### Attribute-Based

Resolve when specific attributes are detected:

| Condition             | Resolves When                |
| --------------------- | ---------------------------- |
| Sentiment = Positive  | Customer seems happy         |
| Issue Type = Question | Informational query answered |
| Priority = Low        | Minor issue handled          |

### Combined Conditions

Combine multiple criteria:

```
Resolved when:
- Agent sent a response AND
- (Customer confirmed OR 24 hours elapsed without follow-up)
```

## Example Resolution Rules

### Customer Confirmed

| Setting       | Value                                                                  |
| ------------- | ---------------------------------------------------------------------- |
| **Name**      | Customer Confirmed Resolution                                          |
| **Type**      | AI-based                                                               |
| **Condition** | Customer explicitly confirms the issue is resolved or expresses thanks |

### Informational Query

| Setting       | Value                                           |
| ------------- | ----------------------------------------------- |
| **Name**      | Information Provided                            |
| **Type**      | Attribute + AI                                  |
| **Condition** | Issue Type = Question AND agent provided answer |

### Time-Based Default

| Setting       | Value                                           |
| ------------- | ----------------------------------------------- |
| **Name**      | No Follow-up (24h)                              |
| **Type**      | Time-based                                      |
| **Condition** | No customer response 24 hours after agent reply |

### Action Completed

| Setting       | Value                                                                        |
| ------------- | ---------------------------------------------------------------------------- |
| **Name**      | Action Successfully Completed                                                |
| **Type**      | AI-based                                                                     |
| **Condition** | Agent successfully performed requested action (refund, password reset, etc.) |

## Resolution Status

Conversations can have these resolution statuses:

| Status         | Meaning                    |
| -------------- | -------------------------- |
| **Resolved**   | Met resolution criteria    |
| **Unresolved** | Did not meet criteria      |
| **Pending**    | Waiting (time-based rules) |
| **Escalated**  | Handed off to human        |

## Impact on Metrics

Resolution rules directly affect:

### Resolution Rate

```
Resolution Rate = Resolved Conversations / Total Conversations × 100
```

### Deflection Rate

```
Deflection Rate = Resolved by AI / Total Conversations × 100
```

### Time to Resolution

```
Time to Resolution = Resolution Timestamp - First Message Timestamp
```

{/* Screenshot: Analytics showing resolution-based metrics */}

## Best Practices

### Be Clear About "Resolved"

Define resolution based on customer outcome, not agent action:

**Good:** Customer confirmed their issue is fixed
**Weak:** Agent sent a response

### Include Multiple Paths

Account for different resolution scenarios:

1. Customer explicitly confirms
2. Time elapses without follow-up
3. Issue type doesn't require confirmation (informational)

### Balance Accuracy vs Coverage

* **Too strict:** Many conversations marked unresolved that were actually successful
* **Too loose:** Resolution rate inflated, doesn't reflect reality

Start with common-sense rules, then refine based on data.

### Review Regularly

Check that resolution status matches reality:

1. Sample resolved conversations — were they actually resolved?
2. Sample unresolved — should some be resolved?
3. Adjust rules based on findings

## Next Steps

<CardGroup cols={2}>
  <Card title="Performance Metrics" icon="chart-line" href="/analytics/performance-metrics">
    Track resolution-based metrics
  </Card>

  <Card title="Tagging Overview" icon="tags" href="/tagging/overview">
    Back to tagging overview
  </Card>
</CardGroup>