Goldilocks Docs
Train

Document Analytics

Track how your documents are being used to answer customer questions

Document analytics help you understand which content is most valuable and identify gaps in your knowledge base.

Retrieval Metrics

Each document tracks how often it's used to answer customer questions.

Retrieval Count

The retrieval count shows how many times a document (or its chunks) was included in AI context when generating responses.

  • Higher counts indicate frequently used content
  • Zero counts may indicate unused or poorly matched content

Last Retrieved

Shows when the document was last used to answer a question. Useful for identifying:

  • Evergreen content (used consistently)
  • Seasonal content (used in bursts)
  • Potentially outdated content (not used recently)

Viewing Analytics

Document List View

On the documents list, you can see:

  • Retrieval count for each document
  • Last retrieved date
  • Trending indicators (if available)

Sort by retrieval count to see your most-used content.

Document Detail View

Click on a document to see detailed analytics:

  • Total retrieval count
  • Retrieval history over time
  • Which chunks are retrieved most often

Using Analytics

Identify High-Value Content

Documents with high retrieval counts are critical to your support. Ensure they are:

  • Accurate and up-to-date
  • Comprehensive
  • Well-written

Find Content Gaps

Low or zero retrieval counts might indicate:

  • Content not matching queries - The AI can't find the content when relevant questions are asked
  • Unused content - Topics customers don't ask about
  • Poor SEO - Content doesn't match how customers phrase questions

Improve Low-Performing Content

If content should be retrieved more often:

  1. Review the document title and content
  2. Add common question phrasings
  3. Include synonyms and variations
  4. Test with the Search feature

Archive Unused Content

If content hasn't been retrieved in months:

  1. Verify it's still accurate
  2. Consider if it's needed
  3. Set to Draft or delete if not valuable

Analytics in the Train Page

The Train page includes document analytics in the Stats Header:

  • Total items and retrieval counts
  • Issues needing attention
  • Unused content filter

Toggle "Show unused only" to focus on content that hasn't been retrieved recently.

For broader analytics, visit the Insights Dashboard →

Best Practices

Regular Reviews

Schedule monthly reviews of document analytics:

  1. Review top 10 most-retrieved documents for accuracy
  2. Check for documents with zero retrievals
  3. Look for trending topics that need more content

A/B Testing Content

If you're unsure which content performs better:

  1. Create two versions of a document
  2. Set one as active
  3. After a period, compare retrieval counts
  4. Keep the better-performing version

Correlate with Outcomes

High retrieval doesn't always mean success. Cross-reference with:

  • Conversation resolution rates
  • Escalation rates
  • Customer feedback

A highly-retrieved document that leads to escalations might need improvement.