Customer service KPIs

The four that matter

• First response time. How long until a person answers. The single best predictor of whether customers feel looked after.
• Resolution time. How long until it is finished. Read the median, not the mean – one three-week case ruins an average.
• Clearance rate. Cleared divided by received, over the same period. Below 100 per cent for several weeks means the backlog is growing and no individual metric will show it.
• SLA rate. The share inside the agreed time. Only meaningful if the agreement is realistic.

Read the median, not the average

Support data is skewed by nature: most cases are quick and a few are enormous. The mean describes neither. The median tells you what a typical customer experienced, which is the question you were asking.

Metrics that mislead

• Tickets per agent. Rewards closing fast, which rewards closing badly. The reopened ticket does not show up.
• Total volume. Rising volume can mean a growing business or a broken product. On its own it says nothing.
• First contact resolution, unless defined very carefully. It is trivially improved by asking fewer clarifying questions.

Count honestly

The most common quiet distortion is counting a resolution on the day it was closed rather than the day it arrived. That makes a catch-up week look like a great week. Attribute the work to the day the request came in and the picture stops flattering you.

How often to look

Weekly for the clearance rate – it is the early warning. Monthly for the rest. Daily reporting in support produces reactions to noise.

What to do with them

A metric that never changes a decision should be deleted. If the SLA rate drops, the question is staffing or target; if resolution time rises, the question is complexity or knowledge. If nobody asks a question, stop producing the number.

How to act on slow response times is in improving response times.