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

# Hotspots

> Identify the topics and issues that matter most to your users—ranked by priority score.

**Hotspots** surfaces the topics and tags that appear most frequently and with the highest impact in your agent's conversations. Instead of reading hundreds of transcripts, Hotspots gives you a ranked view of what your users are actually talking about—and which topics deserve the most attention.

<Note>
  Hotspots is a **Beta** feature. Rankings and scoring logic are actively being refined.
</Note>

## How it works

Every conversation your agent has is automatically classified by topic and tagged. Hotspots aggregates this data across a selected time period and computes a **Priority Score** for each topic based on multiple signals:

| Signal                   | What it measures                                                       |
| ------------------------ | ---------------------------------------------------------------------- |
| **Conversation volume**  | How often this topic comes up                                          |
| **Negative sentiment %** | What share of conversations on this topic have negative user sentiment |
| **Escalation rate**      | How often this topic leads to a human escalation                       |
| **Repeat rate**          | How often users come back with the same topic                          |
| **Avg CSAT**             | Average satisfaction score for conversations on this topic             |

The Priority Score is **normalized from 0 to 100** and used to assign each topic a tier:

* **High** — needs immediate attention
* **Medium** — worth monitoring and improving
* **Low** — stable, no urgent action needed

## Views: Topics vs Tags

Use the toggle at the top to switch between two perspectives:

### Topics view

Shows high-level topics (e.g. `Returns & Refunds`, `Billing Questions`, `Account Access`). Click on any topic to expand it and see the **subtopics** it contains. This gives you a two-level breakdown of the issue.

### Tags view

Shows conversation tags. If your team uses tags to categorize conversations manually or via automation, this view aggregates them by priority the same way topics are aggregated.

## Using date range filters

Select the time period for analysis using the date picker. You can compare different periods to see if certain topics are growing or shrinking in volume.

<Tip>
  Set the date range to the **last 30 days** for an operational view, or to **quarter-to-date** for a strategic review with your team.
</Tip>

## Drilling down into a topic

Click on any topic row to be taken directly to the **Evaluations** view pre-filtered by that topic and date range. There, you can read the actual conversations and identify exactly what's causing the problem.

## Common actions after reviewing Hotspots

* **High-volume topic with high escalation rate** → Add knowledge base articles or update agent instructions for that topic
* **Recurring topic with low CSAT** → Review conversations in Evaluations to find where the agent is falling short
* **Unexpected new topic appearing** → Check if there's a product change or external event driving it