Post-Call Analytics

Outcomes. Drivers.
Why.

Our team turns every call your managed AI voice agents handle — outcomes, sentiment trajectories, conversion drivers, drop-off causes — into clear, board-ready performance reports.

Example dashboard view — names shown are placeholders. Actual results vary by deployment, market, and call mix.

Outcomes · last 30 days
calls graded
73%
resolved without human
Mon
Tue
Wed
Thu
Fri
Sat
Sun
ResolvedTransferredEscalatedAbandoned
Sentiment trajectory
avg per-call start → end
+0.34
↑ 0.08 vs. last month
00:00end
Avg call duration
3:18
↓ 0:22 vs. last month
Containment
73%
↑ 4 pts
Conversion
28%
↑ 6 pts
Cost / resolution
£0.68
↓ £0.14
Drivers

The why,
not just the what.

Our team clusters call reasons semantically rather than by keyword, so the report shows why people are really calling — useful whether your agents handle healthcare intake or lead qualification — and surfaces new themes within days of them emerging.

Top reasons callers reach out
auto-clustered · last 30 days
Repeat prescription32.4%
Reschedule appointment21.1%
Insurance / billing14.8%
New patient intake10.2%
Prescriber update8.0%
Pharmacy hours / location5.6%
Other7.9%
Why calls don't resolve
unresolved calls
Caller asked for a human38.2%
Tool / API timeout22.7%
Out-of-scope intent16.1%
Identity verification failed11.0%
Connection dropped7.2%
Other4.8%
Suggested fix: 60% of “tool / API timeout” maps to lookupPatient(). Investigate EMIS connector.
Slice it however

Pivot by agent, intent,
language, hour, anything.

We can break every metric down the way you need it — by agent version, by intent, by caller's region or by time of day — and shape the reporting around the questions your team is asking.

Containment by agent version & intent
heatmap · 30-day window
All agentsv22 vs v18Top 6 intents
Repeat-Rx
Reschedule
Billing
Intake
Update
Hours
Owen · v22
88%
82%
61%
74%
58%
91%
Owen · v18
71%
65%
42%
55%
38%
72%
Mira · v18
62%
70%
48%
64%
53%
76%
Warehouse-native

Your data,
in your warehouse.

Convexa streams the raw events — calls, turns, function invocations, scores, transcripts — into your Snowflake, BigQuery, Databricks, or Redshift. No ETL, no daily dumps, real continuous sync.

We can deliver the call data into your own warehouse and retain it for as long as you need, so the source of truth stays in the systems your data team already trusts.

-- in your warehouse, today SELECT date_trunc('day', started_at) AS day, agent_version, intent, avg(sentiment_end - sentiment_start) AS recovery, avg(duration_seconds) / 60.0 AS mins, sum(case when outcome = 'resolved' then 1 else 0 end) / count(*)::float AS containment FROM convexa.calls WHERE started_at >= current_date - 30 GROUP BY 1, 2, 3 ORDER BY day DESC;
Built to scale,
human by design.

Post-call analytics and reporting are included as part of every managed Convexa engagement.