Measuring customer experience without vanity metrics

Almost every contact centre has a dashboard full of green. Service level met, handling time down, satisfaction steady. And still customers call twice, send an email afterwards and finally complain about something no tile shows.

That is rarely down to wrong numbers. It is down to numbers that measure something other than what people think. This post names them and shows what works instead.

#Three questions for every metric

You cannot spot a vanity metric by its name, only by its behaviour. Three questions help:

1. Can it improve without customers being better off? If so, sooner or later it will be improved in exactly that way.
2. Does a change trigger an action? A number nobody reacts to is decoration.
3. Can it be traced to a cause? If you do not know why a number rises, you can only hope it falls again.

A metric that answers all three questions well is rare. Usually you need a pair: one number that shows the outcome and one that prevents it from being gamed.

#Five metrics that often mislead

All five have one thing in common: they measure activity, not outcome. They are not wrong, but on their own they are dangerous.

#Metrics that carry decisions

Metrics that hold measure what reaches the customer. Four have proven themselves in many teams:

• First contact resolution: was the request settled in the first contact, across all channels?
• Repeat contact: does the same person get in touch again within a few days about the same topic?
• Effort to resolution: how many contacts and channel switches did a request take?
• Handovers with context: did the customer have to tell the whole story again after a transfer?

All four require contacts to be assigned to a request, not just to a channel. That is the real work. Reporting and analytics on a shared queue and history make it possible.

#From number to cause

A metric tells you that something is happening. Why it happens is in the conversations. Conversation analytics surfaces topics and root causes from transcripts, in aggregate and over weeks.

Journey analytics adds the other side: where on the website or in the app did a request begin, and at which point did people reach for the phone? Together that gives an explanation rather than a guess.

#What AI changes about measurement

AI agents bring a new temptation: the share of conversations that end without a person. It rises quickly and says little. A conversation can end without handover because the request was solved, or because the customer gave up.

The more honest question is the same as for people: was the request settled, and does the person get in touch again? Add the quality of the handover. When an AI agent transfers, the agent should find the context rather than ask again.

The same goes for Univerx Assist. What counts is not how many suggestions were shown, but whether they were used and the request was then resolved.

#People are not a metric

Many vanity metrics arise when team figures are broken down to individuals. Anyone measured personally on handling time ends calls sooner. The result is a nice number and a second call.

In Germany, co-determination comes on top.

Under section 87(1) no. 6 of the Works Constitution Act, technical systems capable of monitoring employee behaviour or performance are subject to the works council's co-determination. Team-level analytics and jointly agreed rules are therefore not only fairer but also more durable.

#A set of metrics that holds

Less is more here. A small set discussed every week has more effect than fifty tiles nobody looks at.

Each of these numbers has an owner and a fixed question: what do we do when it moves? Without that question, even a good metric turns into a shop window.

#How to switch

1. Take stock: which metrics exist today, and who reacts to which?
2. Test: ask the three questions of every metric and answer honestly.
3. Pair: show activity numbers only together with an outcome number.
4. Assign: map contacts to requests, not only to channels.
5. Discuss: a fixed, short weekly format with decisions rather than reports.

Why assigning contacts to requests changes so much is the subject of Journey first, not channel first. How to discuss analytics with the works council is covered in Talking to the works council about analytics.