·
6 min read
Why we measure resolution, not deflection
Deflection counts the conversations you avoided. Resolution counts the problems that stopped existing. Only one of those shows up in your retention numbers.
Two metrics that look alike and aren’t
For a decade, support automation sold itself on deflection: the share of inbound requests that never reached a person. It’s a tidy number, and it’s almost meaningless. A customer who gives up is deflected. A customer who finds a workaround is deflected. A customer who churns quietly three weeks later was, technically, deflected.
Resolution asks a harder question. Did the thing the customer needed actually happen? Was the refund issued, the shipment rebooked, the account unlocked? That requires the agent to take the action, confirm it, and know the difference between “I’ve explained how to do this” and “it’s done”.
What changes when you optimise for it
Teams that switch their north star from deflection to resolution tend to make three moves. They give agents permission to act inside real systems. They start sampling outcomes a week later, not at the end of the chat. And they stop celebrating a quiet queue, because a quiet queue is only good news if the problems went away with it.