A rejected transfer should return the caller to a known conversation state. The implementation must distinguish a failed staff connection from the end of the caller’s own call.
Rejection is a transfer state, not a hangup instruction
Transfer rejection occurs when the destination declines, is busy, or cannot accept the offered conversation. These outcomes may arrive through different signals. Normalise them into application states while preserving the original provider event for investigation.
A proposed attended-transfer workflow keeps the caller connected while a separate staff leg is offered. The agent releases its caller connection only after the application has evidence that the destination accepted and the bridge is established. Exact control depends on the telephone service and must be tested for the selected transfer mode.
A blind transfer may surrender the original connection before the destination answers. Choose that mode only when its failure experience meets the business requirement. RFC 3515 defines the SIP REFER method and outcome notification mechanism; it does not establish the behaviour of a particular carrier configuration.
Test refusal while the caller keeps talking
Consider a fictional caller asking for the service desk. A staff member rejects the offer while the caller says, “Actually, I can wait until tomorrow.” The assistant should consume the rejection, cancel stale transfer speech, and acknowledge the updated preference. It should not restart ringing because an older timer fires.
Use a transfer attempt identifier to connect timers and events to the correct offer. When an attempt reaches a terminal state, later events must not revive it. Retain the caller’s current preference separately from transfer transport status.
| Injected event | Required observation |
|---|---|
| Staff explicitly declines | Caller hears the fallback offer |
| Busy response | No false statement that staff answered |
| Caller hangs up during ringing | Staff offer is cancelled or reconciled |
| Late acceptance after timeout | No unexpected connection to another call |
Verify audio as well as event logs. The caller may technically remain connected while hearing silence, overlapping announcements, or staff-only audio. Each is a different user-facing failure.
Select one fallback owner
Offer an approved queue, alternate destination, or callback request according to staffing policy. Do not cycle indefinitely through the same unavailable person. If the caller declines all alternatives, close the interaction clearly and preserve any actionable request.
Maintaining an extra call leg can increase cost and state-management work. That tradeoff buys recovery control, rather than a guarantee that someone will answer. The related platform work establishes integration experience; this recovery design still requires its own provider tests. Also review transfer timeouts.
Transfer rejection FAQ
Is a busy line the same as a rejected call?
They can share a fallback, but record them separately because staffing and routing fixes differ.
Should a rejected transfer be retried immediately?
Usually offer the caller a choice first. Repeated unsolicited ringing can frustrate both staff and callers.
What proves that recovery worked?
The caller remains audible, hears an accurate explanation, and can select a next step without redialling.