Skills CX operations
Measuring article effectiveness
View counts are the most dangerous metric in self-service. A high-traffic article that precedes a contact is a failure, not a success — customers read it, did not get what they needed, and wrote in anyway. Teams celebrate the traffic; the contact rate tells the truth.
npx rulebase-skills install cx-article-effectivenessWhen to use it
Reach for this when someone says any of these — they are the phrases the skill itself triggers on:
- “are our help articles working”
How it works
The method, in the order the skill runs it. The full procedure — tables, worked examples and the edge cases — is in the skill itself.
Views are not resolution
An article can have high views because it ranks well, because the title promises something it does not deliver, because agents link it habitually, or because the bot surfaces it before handoff. None of those imply resolution.
Core effectiveness signals
Build analysis from conversation exports linked to article events. Exact event names vary by platform; the logic does not.
Ranking articles: useful vs harmful
Sort findings by estimated contact volume attributable to failure, not by view count. A leaky bucket on a high-volume driver beats a dead weight article every time.
AI grounding and view-count misuse
Teams grounding AI agents in "top viewed" articles replicate the view-count mistake at scale. An AI agent citing a high-CAV article automates the path from unhelpful content to confident wrong answers.
Analysis procedure
If you cannot link views to contacts (anonymous help centre, no session stitching), say so explicitly. Report findability and content-quality findings from samples instead of pretending CAV exists.
Related skills
Free and open source, and vendor-neutral — it reads the conversations from whichever helpdesk you already run. Browse all 149 skills · connect your helpdesk over MCP · source on GitHub
