A call ending without a saved record may be normal, abandoned, or a silent failure. Monitor the expected outcome for that interaction before deciding that every recordless call needs the same response.
Missing outcomes need an intent-aware definition
A missing saved outcome occurs when the workflow should have produced a durable result but no matching result can be established. Informational calls, wrong numbers, caller-declined intake, and completed transfers may legitimately have no new customer record.
In a proposed monitoring design, retain a minimal interaction disposition and operation reference. After the call ends, reconcile expected writes against the authoritative system. Use a grace period for legitimate asynchronous work rather than flagging every delayed save immediately.
Classify unresolved cases by the evidence available: no confirmation, validation failure, write rejected, write uncertain, or result missing despite a success claim. These categories point to different operational responses.
Reconstruct a fictional silent failure
At 10:00, a caller confirms a service request. At 10:01, the agent says it was saved. At 10:02, the call ends. The backend record is absent because an adapter swallowed a write error.
The monitor should flag the contradiction between the success claim and missing authoritative result. It should preserve the interaction and operation references for review. Automatically repeating the write without reconciliation could be unsafe if the lookup itself is delayed or incomplete.
| Call disposition | Monitoring response |
|---|---|
| Information only | No missing-write alert |
| Caller declined saving | Record the disposition, not a failed intake |
| Confirmed request, no result | Reconcile and create a review item |
| Write pending asynchronously | Recheck within the defined policy |
| Success claim, rejected write | High-priority correctness defect |
Use fictional incident fixtures to test the monitor. Include both true failures and legitimate no-record calls so reducing missed failures does not flood staff with false alarms.
Assign the unresolved case to someone
A dashboard counter does not recover the caller’s request. Create an owned review item with the safe summary, known action status, contact preference, and next step. Avoid promising a callback unless the business process supports it and the caller’s preference permits it.
Monitoring increases retained operational metadata and review workload. Minimise sensitive content and group duplicate signals for the same interaction. Track whether review items reach a genuine disposition, not merely whether alerts were delivered.
This design is proposed. The HVAC operations case study describes staff-visible requests and follow-up tracking. See failed-transfer queue design for ownership mechanics that also apply here.
Missing-outcome FAQ
Should every call create a customer record?
No. Define legitimate informational and declined-intake dispositions explicitly.
Can the monitor recreate the request automatically?
Only under a safe, authorised reconciliation policy with duplicate protection and valid current intent.
Which cases deserve urgent review?
Prioritise unsupported success claims and confirmed customer requests whose authoritative outcome remains missing or uncertain.