What is a warm transfer?
A warm transfer is the handover of a live call to a person, with the context of the conversation moving across at the same time. The receiving agent arrives knowing who the caller is and what has already been established, so the caller does not begin again. It is distinguished from a cold transfer, where the call arrives with nothing attached to it.
Also called: Attended transfer
Why the handover is where a voice deployment is actually judged
The calls that reach a human are the ones the automation could not resolve, which makes them the calls an organisation can least afford to handle badly. A caller who has spent four minutes with an agent and then has to repeat everything concludes the whole arrangement wasted their time, and they are right. Contact centre leaders judge a voice deployment on this moment more than on containment rate, because it is the moment their own staff experience. What must move across is the established facts and the transfer reason, in a form the receiving agent can read in seconds.
- A caller asked to repeat themselves judges the entire deployment by that moment.
- Containment rate improves while handover quality determines satisfaction.
- Receiving agents need context in the seconds before they greet the caller.
- The transfer reason is what makes routing to the right queue possible.
How a warm transfer works
- The agent determines that a transfer is required and records the reason.
- A context package is assembled from the established facts and the summary so far.
- Routing selects a queue or a named person from the reason given.
- The context package is presented to the receiving agent before the audio connects.
- Audio is bridged once the receiving agent has accepted the call.
- The full exchange is retained as one record across both legs of the call.
How a warm transfer is evaluated
A warm transfer is judged by what the receiving agent has to ask again. Measure the following on transferred calls:
- share of transfers where the caller repeats information already given
- time from transfer request to the receiving agent speaking
- accuracy of the context package against the recording
- transfer abandonment rate while the caller waits
- proportion of transfers reaching the correct queue first time
Limitations, risks and common failure modes
- Context arrives after the audio, so the agent greets the caller unprepared.
- The context package summarises the call wrongly and the agent trusts it.
- Transfers are measured as a failure of containment, so nobody owns their quality.
- A dropped transfer loses the call record for the first leg entirely.