A knowledge base is not a "publish and forget" asset. Like any support tool, it needs to be measured, reviewed, and improved. The challenge is knowing which numbers actually tell you whether the knowledge base is doing its job — and which ones are vanity metrics that look meaningful but are not.
This guide covers the metrics that matter, how to interpret them, and what actions they should drive.
The Core Question
Before listing metrics, it helps to establish what "working" means for a knowledge base. There are three things you want it to do:
- Deflect support tickets — answer questions customers would otherwise have submitted as tickets or chat messages
- Answer questions completely — resolve the visitor's issue without them needing to contact support afterward
- Be findable — reach visitors who have the question, whether they arrive from search engines or from within your product
Every metric worth tracking maps to one of these three outcomes.
Ticket Deflection Metrics
Ticket Volume Trend
The coarsest but most directionally useful metric. If your knowledge base is growing in content and traffic while your ticket volume stays flat or decreases, the knowledge base is likely contributing to deflection. If both grow together, you are adding articles but they are not resolving questions before tickets are opened.
To get a cleaner signal: compare ticket volume in the same categories where you have added knowledge base articles. If billing tickets decrease after you publish a series of billing articles, there is a direct link.
Contact Rate
Contact rate = (number of support contacts) / (number of visitors or active users)
A declining contact rate, holding user base constant, means customers are resolving more issues without reaching support. This metric is more meaningful than raw ticket volume because it controls for growth in your customer base.
Knowledge Base Views Before Ticket Submission
If your platform tracks sessions, you can measure what percentage of customers who submitted a ticket had previously visited a knowledge base article. A high percentage of customers viewing articles then submitting tickets anyway suggests your articles are not fully resolving the issue — the visitor read the article and still could not solve the problem.
Article Quality Metrics
Helpfulness Rating
A simple "Was this helpful? Yes / No" prompt at the bottom of each article generates the most actionable quality signal. Track:
- Overall helpfulness rate across all articles (a baseline)
- Per-article helpfulness rate to identify specific articles that need improvement
- Trend over time — if an article's helpfulness rate drops, something changed (the product UI, a related feature, or a common error the article no longer addresses)
A helpfulness rate under 60% for a frequently viewed article is a clear signal that rewriting or expanding is needed.
Comment Themes from "No" Responses
When you allow customers to leave a comment explaining why an article was not helpful, these comments are some of the most valuable content feedback you will ever receive. Review them monthly. Common themes:
- "The steps don't match what I see on my screen" → the article is outdated
- "This doesn't answer my question" → the article title is attracting traffic it cannot serve
- "I did this and it still doesn't work" → the troubleshooting section is incomplete
Time-to-Resolution Indicator
This requires session tracking: how long does a visitor spend on an article before either navigating away (presumably resolved) or clicking the contact/chat CTA (not resolved)? Very short sessions either mean the visitor found the answer instantly (good) or gave up immediately (bad). Context from the helpfulness rating clarifies which.
Findability Metrics
Search Without Results Rate
Every search query entered into your help center search that returns zero results is a customer with a question you have not written an article for — or have written poorly enough that it does not surface. Export this data monthly and treat it as an article backlog.
Many knowledge base platforms export a list of top search queries. The intersection of high-frequency queries with zero results is your highest-priority content gap.
Top Search Queries
The full list of what customers search for on your help center is more useful than any internal brainstorming session about what to write next. It is your customers telling you directly what they need. Review this list monthly.
External Search Traffic
How many visitors arrive at your help center articles from Google or other search engines? A growing share of external traffic means your articles are ranking for relevant queries and attracting new visitors who may not be existing customers yet. Declining external traffic after a site change may indicate a technical SEO issue.
Track this in Google Search Console, segmented by article URL.
Article Page Views
Raw page views tell you which articles are getting attention. Combined with helpfulness ratings, they reveal your highest-priority optimization opportunities:
| Views | Helpfulness | Action |
|---|---|---|
| High | High | Document and replicate for other articles |
| High | Low | Urgent rewrite — high traffic, low resolution |
| Low | High | Promote more prominently or improve title for search |
| Low | Low | Evaluate whether the article should exist at all |
Operational Metrics
Articles Published vs. Questions Unanswered
Track how many articles you publish each month and compare it against the volume of new questions emerging from chat and tickets. The goal is to shrink the gap — to cover new question categories faster than they accumulate.
Article Staleness
Define a staleness threshold — for example, any article not reviewed in 6 months is flagged. Run a monthly report of stale articles and assign them for review. An article that is technically accurate but covers a UI that changed 8 months ago is doing more harm than good.
Time to First Article (for new topics)
When a new product feature launches, how long does it take for a knowledge base article covering that feature to be published? If the answer is consistently "weeks" or "after the first 50 tickets have arrived", that is a process problem — documentation is not integrated into the product release workflow.
Building a Monthly Review Habit
The metrics above are only useful if someone looks at them regularly and acts on them. A practical monthly review covers:
- Articles with a helpfulness rate below threshold — assign for rewrite
- Top search queries with zero results — add to the content backlog
- Ticket volume by category — identify any category trending up that might indicate a knowledge gap
- External search traffic change — flag drops that might indicate a technical issue
- Stale articles flagged this month — assign for review
This review can be completed in under an hour. The compounding effect over 12 months is a knowledge base that consistently improves rather than slowly degrading.
How Nura24 Supports Knowledge Base Analytics
Nura24 records article views and helpfulness ratings for every published article, accessible from the knowledge base management panel. The AI gap analysis feature automatically identifies questions submitted via live chat and tickets that the knowledge base could not answer, generating a prioritized list of missing articles. External search performance is trackable via standard sitemap and canonical URL support, compatible with Google Search Console. For teams that want a knowledge base that improves automatically over time rather than requiring manual audits, Nura24's connected analytics loop — from ticket question to gap identification to article suggestion — provides a practical foundation.