Analytics and quality

Resolution, not deflection

Fidiora analytics separate deflection from real resolution. A deflected conversation never reached a human, which includes customers who simply gave up. A resolved conversation ended with the customer helped. The two can move in opposite directions, so the number to manage is resolution rate. See deflection vs. resolution and containment rate vs. resolution rate.

The core metrics

Fidiora analytics cover resolution rate, response time, and deflection versus real resolution. The other measures below come from your weekly review and, where you use one, your helpdesk.

• Resolution rate. The share of conversations Fidiora closes without a human, under the resolution definition. It should climb steadily as your docs improve.
• Handoff rate and reasons. How often Fidiora hands off, and which trigger fired: request for a human, low confidence, sensitive topic, frustration, VIP rule.
• Response time. Time to first reply, for AI and for humans after a handoff. See first response time.
• Handoff quality. How often agents ask customers to repeat information, and time from handoff to first human reply.
• Customer satisfaction. CSAT on AI-resolved and on handed-off conversations, tracked separately.
• Reopens. Resolved conversations that came back. A rising reopen rate suggests incomplete answers.

Read handoffs the right way

A high handoff rate is not bad by itself. Handoffs on sensitive or undocumented questions are the system working. Handoffs on questions your docs already cover point to a content or retrieval problem: the article is unclear, split across pages, or uses different words from your customers. Group handoffs by reason each week and you will know exactly where to act.

A weekly quality review

1. 01

   Sample AI-resolved conversations

   Read a sample of conversations Fidiora resolved this week. Check each for a correct, complete answer, the right cited source, and your tone.

2. 02

   Read flagged handoffs

   Read handoffs triggered by frustration or repeated failure. Ask whether the handoff should have happened earlier.

3. 03

   List the content gaps

   Group low-confidence handoffs by topic. The largest groups are the help articles to write next.

4. 04

   Fix and re-sync

   Correct or remove the sources behind any wrong answers, write the missing articles, and re-sync your knowledge sources.

5. 05

   Record what changed

   Note the fixes and watch next week's numbers on those topics. This is how you know the loop is working.

Report to the business

Leadership usually wants three numbers: resolutions this month, cost per resolution, and the effect on the team's queue. Add satisfaction on AI-resolved conversations so the story is about customers helped, not tickets avoided. If you want a template, see support KPI dashboard and how to measure AI support quality.