> ## Documentation Index
> Fetch the complete documentation index at: https://docs.hearflow.io/llms.txt
> Use this file to discover all available pages before exploring further.
# Product Insights Guide
> Learn how to use the Product Insights dashboard to discover patterns in customer conversations and transform them into actionable product feedback.
Discover patterns in customer conversations and transform them into actionable product feedback.
## Overview
The Product Insights dashboard surfaces recurring insights from customer calls, automatically extracted and grouped by AI. Each insight represents a pattern that appears across multiple conversations, backed by verbatim customer quotes as evidence.
**Key Principle** — Every insight is defensible with real customer language. Accuracy is prioritized over coverage — it's better to miss an insight than hallucinate one.
Every insight backed by verbatim customer quotes.
Similar ideas clustered together automatically.
See how often each pattern appears across calls.
## Understanding Insights
### What is an Insight?
An **Insight** represents a recurring idea that appears across customer conversations. Each insight answers: *"What keeps coming up in customer conversations?"*
| Field | Description |
| --------------- | ----------------------------------------------------------------------------------- |
| Title | A clear, stable summary of the recurring idea |
| Category | The type of signal (see Categories below) |
| Product Area | Optional organizational grouping (user-controlled) |
| Priority | Optional triage priority (`p0`, `p1`, `p2`); `p0` is highest. Display-only. |
| Workflow Status | User-facing triage state (`backlog`, `to_do`, `in_progress`, `done`). Display-only. |
| Frequency | How many calls mentioned this insight |
| Last Updated | When the insight was last mentioned |
| Evidence | Verbatim quotes from source calls |
### How Insights Are Created
When a call transcript is processed, the AI extracts potential insights.
Similar ideas are automatically grouped together using semantic similarity.
Each mention links the insight to the source call with a verbatim quote.
As more calls mention the same idea, the insight's frequency increases.
Insights are durable. As more calls come in, insights strengthen — they don't fragment into duplicates.
## Categories
Categories classify the **type of feedback**, not where it belongs in your product. They answer: *"What kind of signal is this?"*
* **Bug** — Something is broken, not working as expected, or producing errors.
* **Feature Request** — Customer explicitly wants new functionality that doesn't exist.
* **Deal Risk** — RARE: Explicit renewal uncertainty, competitive switch, or leadership intervention.
* **Pricing Feedback** — Specific comments about cost, value, or pricing structure.
* **Usability Issue** — A feature exists but is hard to use, configure, or understand.
* **Workflow Gap** — A step, process, or integration is missing entirely.
* **Pain Point** — General frustration when no other category clearly applies.
**Deal Risk Guidelines** — Only categorized as **Deal Risk** when customers explicitly mention:
* Renewal uncertainty: *"we're reviewing whether to renew"*
* Competitive switch: *"we're evaluating alternatives"*
* Leadership intervention: *"our CFO is doing a cost review"*
General complaints or frustrations do **not** qualify as Deal Risk.
## Dashboard Features
### View Modes
Switch between two display modes:
* **Table View** (default): Compact list with sortable columns
* **Grid View**: Card-based layout for visual scanning
### Filtering & Search
| Filter | Description |
| --------------- | -------------------------------------------------------- |
| Search | Full-text search across title, description, and category |
| Category Filter | Click category chips to show only matching insights |
| Product Area | Filter by product area (if configured) |
| Sort By | Order by Frequency or Last Updated |
| Sort Order | Ascending or descending |
Combine filters to narrow down results. Click "Clear filters" to reset.
### Exporting Data
Click the **Export** button to download insights as CSV. The export includes:
* Insight title and description
* Category and product area
* Frequency count
* First and last seen dates
* All linked evidence (quotes and sources)
## Insight Details Panel
Click any insight to open the detail panel with three tabs:
### Evidence Tab
View all verbatim quotes supporting this insight:
* **Quote**: The exact customer language (never edited or paraphrased)
* **Speaker**: Who said it (when available)
* **Sentiment**: Positive, neutral, or negative tone
* **Confidence Score**: AI certainty (percentage)
* **Source Call**: Link to the original conversation
### Related Tab
Discover connections:
* **Timeline**: When the insight was first and last seen
* **Related Insights**: Similar insights based on semantic similarity (shown with match percentage)
### Impact Tab
Understand business context:
* **Affected Accounts**: Which CRM accounts mentioned this insight
* **ARR Impact**: Total annual revenue from affected accounts
* **Sentiment by Account**: Whether each account's mentions were positive, negative, or neutral
## Product Areas
Product Areas represent how **you** organize your product, not how the AI categorizes feedback. They're optional and fully user-controlled.
| | |
| ------------ | ------------------------------------------------ |
| Source | Defined by users in Settings only |
| Assignment | Optional for each insight |
| AI Behavior | Never auto-creates new product areas |
| User Control | Can be assigned, changed, or removed at any time |
**Examples**: Billing, Onboarding, Reporting, API, Integrations. To configure product areas, go to [Settings → Product Areas](https://hearflow.io/settings).
## Best Practices
**Review High-Frequency Insights First** — Insights with higher frequency represent patterns mentioned across many calls. These often indicate widespread customer pain, commonly requested features, or critical usability issues.
**Track Insight Evolution** — Use the timeline feature to understand how long an issue has persisted, whether mentions are increasing or decreasing, and when patterns first emerged.
**Cross-Reference with ARR Impact** — The Impact tab helps prioritize: High-ARR accounts with negative sentiment = urgent. Multiple enterprise accounts mentioning the same issue = strategic importance.
**Use Product Areas for Organization** — As insights accumulate, product areas help with team ownership and routing, sprint planning focus, and stakeholder reporting.
## Troubleshooting
1. Verify integrations are connected in [Settings → Integrations](https://hearflow.io/settings/integrations)
2. Check that calls have been processed in [Dashboard → Processing](https://hearflow.io/dashboard)
3. Ensure calls contain product feedback (general chat won't generate insights)
If you see apparent duplicates:
1. They may have different categories (intentional separation)
2. Post-processing deduplication runs asynchronously — wait a few minutes
3. Report persistent duplicates to the product team for threshold tuning
Evidence is only created when:
1. A verbatim quote is extractable from the transcript
2. The quote meets quality thresholds
3. Citation can be validated against source
If an insight has fewer quotes than expected, the AI may not have found clear, extractable evidence in some calls.
Connect your first data source and start discovering insights from customer conversations.