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

# Find the root cause of a failing call

> A step-by-step guide to performing a deep-dive analysis of a single call. Learn how to use the Call Details view to uncover the root cause of an issue and identify opportunities for improvement.

## The Goal of a Root Cause Analysis

The goal of a root cause analysis is to move beyond simply identifying that a problem occurred and to understand *why* it occurred. By uncovering the root cause, you can implement a lasting solution that prevents the issue from happening again.

## When to Perform a Root Cause Analysis

You should perform a root cause analysis when:

* A critical red flag is triggered (e.g., hallucination, system prompt compliance failure).
* A user reports a poor experience.
* You notice a recurring issue in your Call Logs.

## How to Perform a Root Cause Analysis

<Steps>
  <Step title="Open the Call Details View">
    From the **Call Logs** or **Recent Red Flags**, click on a call to open the **Call Details** view.
  </Step>

  <Step title="Review the Transcript">
    Read the full conversation to understand the context. Pay close attention to the moments leading up to the issue.
  </Step>

  <Step title="Analyze the Events and Metrics">
    Use the **Events** and **Metrics** tabs to get a more detailed view. Look for anomalies like long silences, high crosstalk, or unexpected events.
  </Step>

  <Step title="Form a Hypothesis">
    Based on your review, form a hypothesis about the root cause. For example: "The agent hallucinated because the prompt was ambiguous," or "The agent failed to follow the system prompt because a specific instruction was missing."
  </Step>

  <Step title="Validate Your Hypothesis">
    Look for other calls with similar issues to see if your hypothesis holds true. If it does, you've likely found the root cause.
  </Step>
</Steps>

<img src="https://mintcdn.com/tuner/tFS1DAQs-KSnIeEa/images/call_details.png?fit=max&auto=format&n=tFS1DAQs-KSnIeEa&q=85&s=39bb02a5816e32d486d138b83b61e6ab" alt="Tuner Call Details View" width="1024" height="632" data-path="images/call_details.png" />

## Common Root Causes

<CardGroup cols={3}>
  <Card title="Ambiguous Prompts" icon="circle-question" iconType="solid">
    The agent's instructions were not clear, leading to unexpected behavior.
  </Card>

  <Card title="NLU/ASR Errors" icon="microphone-slash" iconType="solid">
    The agent misheard or misunderstood the user.
  </Card>

  <Card title="Integration Issues" icon="plug" iconType="solid">
    A problem with a backend system or API caused the agent to fail.
  </Card>
</CardGroup>

***

### Diagnose Across All Calls

Manual root cause analysis is the right tool for investigating a single call in depth. When you want to find the root cause of a pattern across many calls at once — for example, why a specific intent keeps failing, or which eval is consistently failing on your worst calls — use the MCP Diagnose agent. It reads your call logs, red flags, and eval results in bulk, identifies the pattern, and tells you what to fix.

<Card title="Diagnose Your Agent" href="/docs/mcp/diagnose" icon="stethoscope" iconType="solid">
  Analyze patterns across all your calls, find the root cause, and get specific fix recommendations.
</Card>

***

### Next Steps

Now that you know how to analyze and triage issues, it's time to learn how to configure your agent to prevent them in the first place.

<Card title="Configuring Your Agent" href="/docs/getting-started/quickstart" icon="cog" iconType="solid">
  A guide to the agent configuration settings in Tuner.
</Card>