Field notes
Five service desk metrics that mislead
Common KPI definitions that look tidy in a monthly pack yet hide how agents and requesters actually experience the desk.
Average first response time collapses when chat bots send an automated acknowledgement. The metric still has value, but only if you separate bot pings from human first touches. Otherwise leadership celebrates speed while requesters wait for a person.
Resolved within SLA often counts tickets closed after a single reply that asks for more information. Reopens then appear as new work. Track reopen rate beside resolve rate, and define resolve as a state the requester confirmed or that held quiet for an agreed window.
Backlog age averages hide the oldest tickets. A desk can look healthy while a handful of complex change requests sit untouched. Report median age and the count of tickets older than twice your target, not only the mean.
Customer satisfaction surveys sent only on easy password resets inflate scores. Sample across categories, or weight survey invitations so high-effort tickets are not underrepresented.
Portal deflection as a success metric fails when agents quietly create tickets for every chat that started in self-service. Pair deflection with subsequent ticket creation within 48 hours from the same account.