Skills CX operations
Measuring real deflection
Almost every automation dashboard reports containment as.
npx rulebase-skills install cx-deflection-analysisWhen to use it
Reach for this when someone says any of these — they are the phrases the skill itself triggers on:
- “what's our real deflection rate”
- “is the bot working”
- “our containment rate is 70% but tickets haven't dropped”
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.
The four ways the naive number misleads
Silent abandonment. The customer gives up mid-flow and closes the tab. No handoff occurred, so the session is counted as contained. This is the single largest contributor, and it is invisible without a resolution signal.
The four buckets
Classify every bot session into exactly one bucket. This is the analysis.
Data you need
One file with both bot sessions and human contacts, so returns on other channels are visible. A bot-only export cannot measure leakage and will just reproduce the vendor's number.
Usage
Arguments
What containment still cannot tell you
Even a correct containment rate does not establish that automation reduced volume. Contained sessions include ones that would never have been tickets, and that fraction is unobservable in historical data.
Cost, honestly
The standard savings claim is deflected contacts × fully loaded cost per contact. It overstates, usually by a lot, because fully loaded cost includes fixed overhead that does not leave when volume falls.
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
