CitaiCitai
Back to blogCreate free account
ProductMarch 17, 2026·7 min read

KB Health Score: Monitor and Improve Your Knowledge Quality

Citai Team

March 17, 2026 · 7 min read

What is the KB Health Score?

The Health Score is an indicator from 0 to 100 that reflects the quality and completeness of a knowledge base. It doesn’t just measure document count — it evaluates 5 different dimensions that directly impact chatbot response quality.

Health Score Formula: The score is calculated as a weighted average of 5 components: Documents (25%), FAQs (15%), RAG Quality (30%), Knowledge Gaps (15%), and Freshness (15%). Each component is evaluated from 0 to 100 and then weighted.

The 5 Components

25%Documents: quantity and diversity
15%FAQs: frequent question coverage
30%RAG Quality: average confidence
15%Gaps: unanswered questions
15%Freshness: content age

1. Documents (25%)

Evaluates the number of processed documents and diversity of generated chunks:

  • 0-2 documents: Low score (0-30)
  • 3-10 documents: Medium score (30-70)
  • 10+ documents with good diversity: High score (70-100)

2. FAQs (15%)

FAQs are direct answers that avoid LLM calls. This component measures:

  • Active FAQ count relative to the volume of received questions
  • FAQ hit rate — percentage of questions resolved by FAQ match

3. RAG Quality (30%)

The heaviest component. Evaluates the average confidence of responses generated with documents from this KB:

  • Confidence < 0.4: Unreliable responses
  • Confidence 0.4-0.7: Acceptable responses
  • Confidence > 0.7: High-quality responses

4. Knowledge Gaps (15%)

Measures how many questions remain without a satisfactory answer:

  • Proportion of queries with low confidence
  • Questions without relevant chunks
  • Negative feedback received

5. Freshness (15%)

Evaluates how up-to-date the content is:

  • Last document uploaded: Days, weeks, or months ago?
  • Last reprocessing: Have chunks been updated recently?

Score Interpretation

Range Status Meaning
0-40 🔴 Poor KB needs more content or better quality
40-70 🟡 Decent Functional but with room for improvement
70-100 🟢 Excellent Consistently high-quality responses

How to Improve Each Component

Improvement guide by component:
  • Low documents: Upload more relevant documentation. Prioritize quality over quantity — one well-structured document is worth more than 10 scanned ones
  • Low FAQs: Review "Knowledge Gaps" in the analytics dashboard and convert frequent questions into FAQs. Use automatic suggested FAQ generation
  • Low RAG quality: Use the Playground to diagnose. Adjust chunk_size, chunk_overlap, and similarity_threshold. Reprocess documents after adjusting
  • High gaps: Each gap is an opportunity. Add content covering unanswered questions or create specific FAQs
  • Low freshness: Review and update documents periodically. Upload new versions when content changes

Health Score Cache

The Health Score calculation involves multiple database queries. To maintain performance:

  • Redis cache: The score is cached for 1 hour
  • Automatic invalidation: When uploading documents, creating FAQs, or reprocessing chunks, the cache is invalidated
  • On-demand calculation: Recalculated only when the KB is accessed and the cache has expired

Where to See the Health Score

The Health Score appears as a visual badge on each knowledge base card:

  • Dynamic color: Red (0-40), yellow (40-70), green (70-100)
  • Tooltip with details: Hover to see the component breakdown
  • Analytics dashboard: Global view of all KBs with their scores

Conclusion

The KB Health Score transforms knowledge base management from an intuitive task to a data-driven one. Instead of guessing whether your documentation is sufficient, you have a clear indicator with concrete actions to improve. Keep your KBs above 70 to ensure high-quality responses.

Monitor your knowledge base health in real time. Get started with Citai →

Try Citai for free

Create your intelligent knowledge base in minutes. No credit card required.

Create free account

Related articles

Technology

Smart Routing: Automatic Knowledge Base Selection with Embeddings

How Citai automatically selects the best knowledge base for each question using embedding centroids and cosine similarity.

Read article → Product

Intelligent Knowledge Base: A Guide for Businesses

Learn to build an effective knowledge base: documents, FAQs, health score, and smart routing.

Read article → Practical Guide

RAG Playground: how to diagnose and fine-tune your knowledge base

Practical guide to using Citai RAG Playground: pipeline diagnostics, parameter tuning, A/B comparison and batch testing.

Read article →
Smart Routing: Automatic Knowledge Base ... Audit Logging: Complete Traceability for...