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

# Review Audit Results

> Understand scores, find problem calls, and read recommended fixes

Open **View details** from an Audit Report to review its results.

The details page has two tabs:

* **Summary** — overall scores and common issues.
* **Call Audits** — results for individual calls.

## Review the Summary

The Summary can show:

* **Total Calls**
* **Average Score**
* **Calls Below 80% Score**
* **Total Issues**
* Credits used by the audit

It also groups quality into:

* **Callflow Adherence**
* **Function Accuracy**
* **Postcall Accuracy**
* **Transcription Quality**

<Tip>
  Start with **Calls Below 80% Score** and **Critical** issues. These usually need attention first.
</Tip>

## Understand issue severity

| **Severity** | **Suggested action**                                         |
| ------------ | ------------------------------------------------------------ |
| **Critical** | Review immediately. The issue may seriously affect the call. |
| **High**     | Review soon. The issue has a strong effect on quality.       |
| **Medium**   | Review when improving the agent.                             |
| **Low**      | A smaller issue or improvement opportunity.                  |

Select a severity, category, or affected call to see more details.

## Review Call Audits

Open **Call Audits** to see one row for each audited call.

You can filter by:

* Date range.
* Agent version.
* Call duration.
* Disconnect reason.
* Conversion status.

The table shows the overall score and separate scores for Postcall, Transcriber, Callflow, and Function quality.

Select **View issues** to inspect a call.

## Understand an issue

The issue panel may show:

| **Field**           | **What it means**                             |
| ------------------- | --------------------------------------------- |
| **Problem**         | What went wrong.                              |
| **Severity**        | How important the issue is.                   |
| **Proof**           | The call evidence used to identify the issue. |
| **Expected**        | What should have happened.                    |
| **Actual**          | What happened during the call.                |
| **Recommended Fix** | A suggested way to improve the agent.         |

Issues can appear under:

* **Transcription Issues**
* **Function Call Issues**
* **Postcall Variable Issues**
* **Callflow Adherence Issues**
* **System Log Errors**

## Check Extra Variables

If the strategy contains Extra Variables, their values appear in the call issue panel.

Boolean and Issue variables may also appear in the Summary, where you can open the matching calls.

## Check Re-transcription

When available, select **Re-transcription** to compare the call with the improved transcript used during auditing.

This can help you understand whether an issue came from transcription or from the agent's behavior.

## Improve the agent

After reviewing an issue:

1. Read the **Proof**, **Expected**, and **Actual** values.
2. Check the **Recommended Fix**.
3. Open the affected voice agent.
4. Create or edit a safe test version.
5. Update the prompt, function, variable, voice, or transcriber causing the issue.
6. Test the call again before publishing.

<Warning>
  Do not change a live agent based on one call alone. Check whether the same issue appears across multiple calls first.
</Warning>

## Fix issues with AI FDE

Use AI FDE to investigate the finding or prepare a safe fix in a new version.

<CardGroup cols={2}>
  <Card title="Choose an AI FDE mode" icon="toggle-on" href="/ai-fde/choose-a-mode">
    Use Ask to investigate or Plan to approve a fix before building.
  </Card>

  <Card title="Build and update agents" icon="wand-magic-sparkles" href="/ai-fde/build-and-manage-agents">
    Update the affected prompt, function, variable, voice, or transcriber.
  </Card>
</CardGroup>