The dashboard counts 50 transfers, but the office cannot say how many callers actually reached a person. The metric records the AI's attempt rather than the human outcome the business wanted.
For teams evaluating reception, distinguish transfer submission, telephone answer and staff acceptance. They can be separate events with different available evidence.
What is transfer acceptance?
Acceptance means the intended recipient has taken the call or responsibility according to the business's defined workflow. The evidence may differ between a direct telephone transfer and an assisted consultation.
A provider accepting a transfer request is not necessarily a human answer. A telephone answer may be voicemail. A human answering may still decline the requested responsibility.
Define an evidence-based transfer funnel
Use only states the actual provider and application can observe. The proposed funnel below shows where definitions need to remain explicit.
| State | Evidence question |
|---|---|
| Requested | Did the caller ask for the supported handoff? |
| Submitted | Did the application send the transfer action? |
| Routed | Did the provider accept the destination attempt? |
| Answered | What establishes that the destination answered? |
| Accepted | What establishes the intended staff responsibility? |
| Recovered | What owned next step followed an unsuccessful handoff? |
If the system cannot observe a later state reliably, mark it unknown or measure it through an approved additional process. Do not rename submitted as accepted to simplify reporting.
Correlate attempts without inflating outcomes
Several attempts can relate to one caller request. Preserve attempt-level records and a request-level outcome so a retry does not look like another successful customer handoff.
The transfer metadata guide helps connect call legs, while the warm-versus-blind comparison explains why acceptance evidence can differ by mechanism.
Keep sensitive caller information out of broad analytics labels. The report usually needs route, outcome and relevant timing, with controlled access to detailed evidence for investigation.
Use the funnel to improve the route
A high submitted count with low answer evidence points to a different issue from staff answering and declining inappropriate requests. Review staffing, destination mapping and eligibility according to the observed boundary.
My voice platform contribution provides application context, not a claimed transfer-success rate. A new report should state its denominator, observation period and unknown-state treatment.
Inspect a small set of fictional answered, voicemail, declined and no-answer scenarios before using the metric in a client report. The names in the dashboard should match what those cases actually prove.
Is a successful transfer API response proof of human acceptance?
Check the provider's semantics. It may establish only an intermediate routing or submission result.
Should voicemail count as a completed staff handoff?
Define voicemail separately unless the business explicitly treats it as the intended owned outcome.
What should the dashboard show first?
Display attempts, observable acceptance and unknown outcomes distinctly in the voice AI reporting brief.