Article

Customer service metrics:
the eight worth tracking.

By Marija Jovanović

Updated

3 minute read

Most support dashboards track way too many numbers and read none of them together. Eight metrics cover what a support team actually needs to track: how fast it answers, how much effort each answer takes, whether the answer worked, what it costs, and how much of the work never needed a person in the first place. Everything else is a slice of one of these.

Speed: two numbers, not one

First response time is the wait between a customer’s first message and the first reply that addresses it. It is the number customers remember, and the easiest to improve dishonestly, because an automatic acknowledgement can count as a reply unless you exclude it. Report the median rather than the mean: one ticket that waited a weekend drags the mean across the whole month.

Average handle time is the total time one contact takes, including talk or typing time, plus hold, plus the after-contact work, divided by the number of contacts handled. It can fall for a good reason, such as better documentation or a faster way to look up an order, and for a bad one, such as agents closing early to protect the number. Read it with first contact resolution, otherwise a falling handle time looks like progress while customers come back twice.

Outcome: did the answer work?

First contact resolution is the share of resolved tickets that were solved in one contact and stayed closed through an agreed re-open window, commonly 24 to 72 hours. It is the metric that catches a fast, wrong answer, so it belongs next to every speed number above.

Resolution rate is the number of tickets resolved divided by the number of tickets received in the same period. It can pass 100% in a week when the team clears the backlog, and it can sit above 90% every week while the backlog triples under it. Track the backlog beside it; the rate alone will hide it.

Cost: the number that turns the rest into money

Cost per ticket is total support spend, including salaries, tools and management overhead, divided by the number of tickets resolved. Most teams only count agent salaries and undercount it. It is the metric that gives every other metric a price, and the one most often left off the dashboard.

Automation: what never reached a person

Deflection rate is the share of issues fully resolved without a human, out of everything raised in the period. Containment rate is narrower: the share of conversations a bot handled that ended without escalating. A bot that sees a third of the queue and a bot that sees the whole queue can both print a high containment rate; only deflection says what the queue actually lost. Both are gamed the same way, by counting an abandoned conversation as resolved, so define resolved as the customer confirming or not returning, and say so on the dashboard.

Agent assist is the one entry here that is not a number. It is AI that drafts the reply, picks the answer and sums up the thread for a human agent, who still reviews and sends the reply. It moves the metrics above without appearing on any of them, so it is worth knowing which of your numbers it is moving.

How to read the eight together

No one number here is honest by itself. Read them in pairs: first response time with average handle time, first contact resolution with resolution rate, deflection with containment, and cost per ticket with all of them. Each entry in the support metrics glossary shows the formula, a worked example, and how the metric gets inflated without a single extra customer being helped.

FAQ

Which customer service metrics matter most?

Eight cover the job: first response time, average handle time, first contact resolution, resolution rate, cost per ticket, deflection rate, containment rate, and whether agent assist is moving any of them.

Every other metric on a support dashboard is a slice of one of these.

What is the difference between deflection rate and containment rate?

Deflection rate counts every issue resolved without a human, out of all issues raised in the period.

Containment rate counts only the conversations a bot handled and asks how many ended without escalating. A bot can post a high containment rate while deflecting very little of the queue.

Why can resolution rate go above 100%?

Because it divides tickets resolved in a period by tickets received in it, and the resolved work can include older tickets.

A week in which the team clears backlog prints above 100%. Sustained readings above 100% mean the backlog is shrinking.

See what the agent resolves on your own docs.

Connect a knowledge base, ask it the questions your customers ask, and read the answers it gives. Free plan, no card.

Keep reading

Customer Service Metrics: The 8 Worth Tracking | Outlearn