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

# CX Score

> Understand how the CX Score measures conversation quality across five dimensions

The CX Score is a 0-100 quality metric that summarizes how well a conversation went. It combines scores from five dimensions — Resolution Quality, Response Coherence, Tone, Efficiency, and Customer Effort — into a single number you can track over time, filter by, and compare across agents.

## How the Score Is Calculated

Each conversation is scored on five dimensions (0-100 each). The overall CX Score is the weighted average:

| Dimension          | Default Weight | What It Measures                                                 |
| ------------------ | -------------- | ---------------------------------------------------------------- |
| Resolution Quality | 25%            | Did the agent address the user's request?                        |
| Response Coherence | 20%            | Were responses consistent and contradiction-free?                |
| Tone               | 20%            | Did the tone match the agent's configured guidelines?            |
| Efficiency         | 20%            | Was the conversation handled without unnecessary back-and-forth? |
| Customer Effort    | 15%            | How easy was it for the user to get their answer?                |

**Example:** If a conversation scores Resolution Quality: 85, Response Coherence: 90, Tone: 75, Efficiency: 80, Customer Effort: 70, the CX Score is (85×0.25 + 90×0.20 + 75×0.20 + 80×0.20 + 70×0.15) = **81**.

Weights are configurable in the [Thread Evaluator](/features/thread-evaluator) settings. Adjust them to reflect what matters most for your use case — for example, increase Resolution Quality for support agents or Tone & Empathy for customer-facing agents.

<Warning>
  Weight changes apply to new evaluations only. Historical scores are not recalculated.
</Warning>

## Dimension Details

### Resolution Quality (Default: 25%)

Did the agent fully address the user's request within its capabilities?

| Score  | Criteria                                       |
| ------ | ---------------------------------------------- |
| 90-100 | Completely resolved with a thorough answer     |
| 70-89  | Mostly resolved, minor gaps                    |
| 50-69  | Partially resolved, some questions remain      |
| 30-49  | Attempted but failed to resolve the core issue |
| 0-29   | Did not address the user's actual request      |

### Response Coherence (Default: 20%)

Were responses accurate, consistent, and free of contradictions?

| Score  | Criteria                                           |
| ------ | -------------------------------------------------- |
| 90-100 | No contradictions, all responses logically sound   |
| 70-89  | Minor inconsistencies that don't confuse the user  |
| 50-69  | Some contradictions or confusing statements        |
| 30-49  | Clear contradictions between responses             |
| 0-29   | Responses contradict each other or are nonsensical |

<Note>
  The evaluator does not have access to the agent's knowledge base, so it cannot verify factual accuracy against your documents. This dimension focuses on logical consistency within the conversation.
</Note>

### Tone (Default: 20%)

Did the tone match the agent's configured guidelines?

| Score  | Criteria                                           |
| ------ | -------------------------------------------------- |
| 90-100 | Matched the tone specified in agent instructions   |
| 70-89  | Generally aligned, minor deviations                |
| 50-69  | Partially aligned, some inconsistencies            |
| 30-49  | Noticeably misaligned with expected tone           |
| 0-29   | Completely ignored or contradicted tone guidelines |

### Efficiency (Default: 20%)

Was the conversation handled without unnecessary exchanges?

| Score  | Criteria                                     |
| ------ | -------------------------------------------- |
| 90-100 | Resolved in minimum necessary exchanges      |
| 70-89  | Efficient with minor redundancy              |
| 50-69  | Some unnecessary back-and-forth              |
| 30-49  | Significant inefficiency                     |
| 0-29   | Extremely prolonged or circular conversation |

### Customer Effort (Default: 15%)

How easy was it for the user to get their answer?

| Score  | Criteria                                       |
| ------ | ---------------------------------------------- |
| 90-100 | Effortless experience, user guided smoothly    |
| 70-89  | Low effort, minor friction points              |
| 50-69  | Moderate effort required                       |
| 30-49  | High effort, user had to work hard             |
| 0-29   | Extremely difficult, user struggled throughout |

## Interpreting Scores

| Range    | Rating         | What to Do                                         |
| -------- | -------------- | -------------------------------------------------- |
| 90-100   | Exceptional    | Use as a training example                          |
| 80-89    | Good           | Meets quality standards                            |
| 70-79    | Acceptable     | Minor improvements possible                        |
| 60-69    | Needs Work     | Review conversations in this range for patterns    |
| 50-59    | Below Standard | Investigate root cause — check dimension breakdown |
| Below 50 | Critical       | Immediate attention required                       |

<Note>
  Conversations below the minimum validation score (default: 70) or with any dimension below its threshold (default: 50) are automatically flagged for review. Configure thresholds in the [Thread Evaluator](/features/thread-evaluator) settings.
</Note>

## Where CX Scores Appear

### Evaluation Dashboard Tab

The **Evaluation** tab on the [Dashboard](/features/evaluation-dashboard) shows:

* CX Score gauge with overall average
* Dimension averages as progress bars
* Score distribution (Poor / Needs Improvement / Good)
* CX Score trend over time
* CX Score by agent comparison

### Conversation Sidebar

When viewing a conversation in [Chat History](/features/conversations-list), the sidebar shows the CX Score gauge and dimension breakdown. Click **Run Evaluation** to score a conversation manually.

### Filtering by CX Score

In **Chat History**, use the **CX Score Range** filter to find conversations by score range (Poor 0-49, Needs Improvement 50-79, Good 80-100). A **CX Dimension Filter** section lets you set per-dimension thresholds — for example, Customer Effort \< 50 **and** Efficiency ≥ 70.

### CX Score Distribution

View how scores are distributed across conversations:

```
0-49:   ████ 15%
50-79:  ████████████ 45%
80-100: ██████████ 40%
```

### CX Score Trends

Track average scores over time:

* Daily averages
* Weekly trends
* Monthly comparisons
* Before/after improvement initiatives

### CX Score by Agent

Compare performance across agents:

| Agent           | Avg Score | Conversations | Trend |
| --------------- | --------- | ------------- | ----- |
| Support Agent   | 82        | 1,234         | ↑ +3  |
| Sales Agent     | 78        | 567           | → 0   |
| Technical Agent | 85        | 890           | ↑ +5  |

### Dimension Analysis

Identify which dimensions need attention:

```
Resolution Quality:  ████████████████████ 82
Response Coherence:  ██████████████████████ 88
Tone:                ████████████████ 72
Efficiency:          ██████████████████ 78
Customer Effort:     █████████████████ 75
```

In this example, Tone needs improvement.

***

## Filtering by CX Score

### Overall CX Score Filter

In **Chat History**, filter conversations by their overall CX score using the **CX Score Range** filter in the secondary filters panel. Choose from predefined ranges (Poor, Needs Improvement, Good) or use dimension-level filters for finer control.

### Dimension Score Filters

A dedicated **CX Dimension Filter** section lets you set thresholds for each of the five dimensions:

* Choose **\<** (less than) or **≥** (greater than or equal) for each dimension
* Enter a score between 0 and 100
* Combine multiple dimension filters to narrow results (e.g., Customer Effort \< 50 **and** Efficiency ≥ 70)

Only dimensions with a score value entered are applied as filters. Toggle the operator at any time — the filter updates immediately.

## Exporting CX Score Data

When exporting conversations from **Chat History**, CX score details are included automatically. The export contains:

* Overall CX score
* Individual dimension scores (Resolution Quality, Response Coherence, Tone, Efficiency, Customer Effort)

This allows you to analyze CX scores in external tools like spreadsheets or BI dashboards.

***

## Using CX Score Data

**Find training needs:** Filter Chat History by low dimension scores, read 5-10 conversations, and look for patterns. If Resolution Quality is consistently low, check for knowledge gaps. If Tone is low, review the agent's tone instructions.

**Compare agents:** The Evaluation tab's CX Score by Agent chart shows which agents score highest and lowest. Investigate what top performers do differently — it may be their instructions, knowledge sources, or tool configuration.

**Track improvements:** After making a change (updating knowledge, editing instructions, adjusting tools), watch the CX Score trend chart over the following week to confirm the metric moves in the right direction.

## Troubleshooting

<AccordionGroup>
  <Accordion title="Scores seem too high or too low">
    Review dimension weights in [Thread Evaluator](/features/thread-evaluator) settings. If weights don't match your priorities, the overall score may not reflect perceived quality. Also check custom instructions — vague guidelines produce inconsistent evaluations.
  </Accordion>

  <Accordion title="One dimension consistently low">
    Investigate the root cause for that specific dimension. Low Resolution Quality often means missing documentation. Low Tone may indicate the agent's tone instructions need updating. Low Efficiency could mean the agent asks too many clarifying questions.
  </Accordion>

  <Accordion title="Scores vary widely day to day">
    Check if conversation complexity varies by day (e.g., more technical questions on weekdays). With small sample sizes, individual outliers have a big impact. Filter out outliers for trend analysis and focus on weekly averages rather than daily.
  </Accordion>
</AccordionGroup>

## Related Features

* [Thread Evaluator](/features/thread-evaluator) - Configure evaluation settings, weights, and thresholds
* [Evaluation Dashboard](/features/evaluation-dashboard) - View CX Score charts and trends
* [Conversations](/features/conversations-list) - Filter and export conversations by CX Score
* [Diagnose Quality Issues](/guides/diagnose-quality-issues) - Step-by-step guide to investigating low scores