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

# QA and analytics

> Review conversations, analyze dashboards, and use Smart Analyst for insights across hundreds of calls.

Use Dashboards for high-level metrics, Conversation Review for transcript inspection, and Smart Analyst for AI-powered analysis across hundreds of conversations without manual review.

<Tip>
  **Skip manual call reviews with Smart Analyst.** Instead of listening to individual calls, use [Smart Analyst deep sampling](/smart-analyst/introduction) to analyze up to 500 conversations per query. Ask questions like *"What are the top handoff reasons this week?"* and get structured insights in minutes. It's the fastest way to identify patterns and prioritize fixes.
</Tip>

## Quick reference

| I need to...                            | Action                                                                    | Time estimate |
| --------------------------------------- | ------------------------------------------------------------------------- | ------------- |
| Check daily performance                 | Configure > Dashboards → Standard → Review key metrics                    | 5 min         |
| Listen to recent calls                  | Analytics > Conversations > Voice → Filter → Listen                       | 10-20 min     |
| Find knowledge gaps                     | Configure > Dashboards → Unhandled queries                                | 10 min        |
| Analyze trends across hundreds of calls | Analytics > [Smart Analyst](/smart-analyst/introduction) → Ask a question | 5 min         |
| Review a specific topic                 | Analytics > Conversations > Voice → Filter by topic                       | 5 min         |
| Export call data                        | Analytics > Conversations > Voice → Export CSV/JSON                       | 5 min         |
| Tag problematic calls                   | Open call → Add category/annotation                                       | 2 min         |
| Check containment rate                  | Configure > Dashboards → Standard → Containment metric                    | 2 min         |
| Review safety issues                    | Safety dashboard → Check flags                                            | 10 min        |

## Pick the right dashboard

| Use case                                     | Go here                                                                                    |
| -------------------------------------------- | ------------------------------------------------------------------------------------------ |
| Daily health check (calls, containment, AHT) | [Standard dashboard](/analytics/dashboards/standard)                                       |
| Risk or policy compliance                    | [Safety dashboard](/analytics/dashboards/safety)                                           |
| Anything bespoke (e.g. bookings, revenue)    | [Custom dashboards](/analytics/dashboards/custom) – contact support if you need one set up |

Dashboards default to **Live** and **last 7 days**. Change the environment or date range at the top-right.

## Example workflow: did my FAQ edit improve containment?

1. Open the **Standard dashboard**.
2. Set the date range:
   * **From** = the day you promoted the change
   * **To** = today
3. Check **Containment rate**.
   * If it went up → good sign
   * If it dropped or stayed flat, drill down to conversations (next section)

## Review conversations

1. Go to **Analytics > Conversations > Voice**.
2. Use filters:
   * **Date range** – same window you used in the dashboard
   * **Topic** – select the FAQ you edited
3. Click a call to open the transcript.
4. Check:
   * Did the agent serve the new answer?
   * Did the caller stay in self-service or ask for a human?
5. If something looks off, tag it.

### Tagging and annotations

At the bottom-right of a transcript:

* **Category** – pick a label like *Needs KB fix* or *Escalated unnecessarily*
* **Annotation** – flag *Wrong transcription* or *Missing topic*

These tags surface in QA reports and help you spot patterns over time. They are also the primary [feedback loop from human reviewers back into the agent](/call-handoff/introduction#feedback-loops-from-the-human-side) — escalations tagged *Escalated unnecessarily* or *Wrong topic* are where to look first when tightening handoff rules.

## Export transcripts (optional)

Need raw data for deeper analysis?

1. In **Conversations**, select your filters.
2. Click **Export** → *CSV* or *JSON*.
3. You'll get a download link with call metadata and the full transcripts.

See [Call data → Transcripts](/call-data/studio-transcripts) for field definitions.

## Common workflows

### Daily performance check

1. Open Standard dashboard
2. Review key metrics vs. baseline:
   * Call volume
   * Containment rate
   * Average handle time
   * Handoff rate
3. Identify any anomalies
4. Drill into Conversation Review if needed

**Time:** 5-10 minutes

### Weekly quality review

1. **Start with [Smart Analyst](/smart-analyst/introduction)** – run a deep sampling query to surface the week's key patterns: *"What are the top issues and failure reasons from the last 7 days?"*
2. Review Smart Analyst findings and identify areas that need manual review
3. Listen to 5-10 sample calls flagged by Smart Analyst or filtered in Conversation Review
4. Tag issues for follow-up
5. Update knowledge or settings based on insights

**Time:** 20-40 minutes (Smart Analyst reduces manual review time)

### Investigating a spike in handoffs

1. Go to **Configure > Dashboards** → Standard dashboard
2. Identify when handoff rate increased
3. **Ask [Smart Analyst](/smart-analyst/introduction):** *"What are the top reasons for handoffs in the last 3 days? Give me percentage breakdowns."* – deep sampling analyzes up to 500 conversations to surface root causes
4. Review specific conversations in **Analytics > Conversations > Voice** to validate Smart Analyst findings
5. Add missing knowledge or adjust handoff rules based on the patterns identified

**Time:** 15-20 minutes (Smart Analyst replaces manual filtering and pattern identification)

## Related pages

* [Smart Analyst](/smart-analyst/introduction) - AI-powered conversation analysis with deep sampling
* [Performance monitoring](/learn/maintain/performance-monitoring) - Detailed performance analysis
* [Health checks](/learn/maintain/health-checks) - Proactive monitoring routines
* [Standard dashboard](/analytics/dashboards/standard) - Key metrics overview
* [Conversation Review](/analytics/conversations/review) - Transcript inspection
* [Call data](/call-data/introduction) - Accessing raw call data
* [Alerts API](/api-reference/alerts/introduction) - Set up automated alerts for operational metrics

## When to escalate

| Situation                           | Who to contact                         |
| ----------------------------------- | -------------------------------------- |
| Metrics flat after multiple tweaks  | PolyAI support – may need flow changes |
| Consistent ASR errors               | PolyAI support – include call IDs      |
| Dashboard missing a metric you need | PolyAI support                         |
| Safety dashboard shows spikes       | PolyAI support – escalate immediately  |

## Next steps

* Check the [standard dashboard](/analytics/dashboards/standard) now to make sure your latest edit had the intended effect.
* Use [Smart Analyst](/smart-analyst/introduction) to analyze hundreds of conversations at once – its deep sampling capability surfaces patterns, failure reasons, and sentiment trends far faster than manual review.
* Set up [health checks](/learn/maintain/health-checks) for proactive monitoring.
* Configure [automated alerts](/api-reference/alerts/introduction) for latency, errors, and call volume.