Call Analytics & Reporting
Dashboard → Analytics shows what your agent is doing across all calls. This guide covers which numbers matter and which are comfortable but not useful.
The Metrics Worth Watching
- Call volume over time — establishes your baseline and reveals when you are actually busy
- Lead capture rate — calls that produced complete lead details, as a share of genuine enquiries
- Call outcomes — how calls are being classified once they complete
- Score distribution — the spread of lead quality, not just the average
- Call duration — useful mainly as an anomaly detector
- After-hours share — how much of your volume arrives when nobody was available
💡 Tip: After-hours share is the number that most often justifies the whole system. If a third of your calls arrive outside working hours, that is a third you were previously losing.
Reading Duration Correctly
Average call duration is not a quality measure on its own. Long calls can mean thorough qualification or an agent that cannot close. Short calls can mean ruthless efficiency or callers hanging up. Segment by outcome before drawing conclusions.
⚠ Note: A sudden drop in average duration is worth investigating the same day. It usually means calls are failing early rather than that your agent became efficient overnight.
Score Distribution Beats Score Average
An average score tells you very little. The distribution tells you whether you have a steady stream of qualified enquiries with occasional noise, or a flood of low-quality calls with rare exceptions. Those two situations need completely different responses.
Turning Numbers Into Changes
Analytics is only useful if it changes something. The usual chain:
- Notice a pattern in Analytics
- Read ten transcripts from the affected calls in Call Logs
- Identify the specific conversational failure
- Make one prompt or setting change
- Check the same metric a week later
✓ Success: The dashboard tells you where to look. The transcripts tell you what to fix. Skipping the transcripts is how people end up changing settings at random.