Key metrics in the admin dashboard

Resolution rate

The percentage of conversations where the chatbot provided an answer (not a fallback). A resolution rate above 80% is a healthy baseline for a well-maintained knowledge base.

Average confidence score

The mean confidence across all answers in a given period. Rising confidence indicates your knowledge base is becoming more comprehensive. Falling confidence suggests users are asking questions outside your current coverage.

`wasResolved` rate

The percentage of conversations where users gave a thumbs-up. This is the most reliable signal of actual helpfulness because it reflects user judgment, not model confidence. Target: > 70%.

Unanswered questions

Count of exchanges where the fallback message was shown. Drill down to see which topics are generating misses.

Escalations

How many times users clicked your human escalation link. A high escalation rate with high wasResolved means your fallback experience is working well. A high escalation rate with low wasResolved means the knowledge base has significant gaps.

Where to find these metrics

Trend charts for resolution rate and confidence (daily/weekly/monthly)
CSAT (thumbs-up percentage)
Top unanswered question clusters
Per-KB breakdown if you have multiple knowledge bases

How to improve performance

The most effective levers are:

Fill KB gaps — add documentation for the topics appearing in unanswered questions
Improve source quality — rewrite dense or ambiguous documents in clearer prose
Update stale content — outdated documents lower confidence on related questions; re-upload updated versions
Segment by KB — if one product area has low resolution rate, focus content additions there

What data is stored for analytics?

0xDoktor never stores raw user questions. Queries are SHA-256 hashed before being logged. Analytics show patterns and clusters — not the verbatim text of conversations. This design is GDPR-compliant by architecture.

How Do I Measure My 0xDoktor Chatbot's Performance?