The Contact Centre Metrics Worth Managing (and the Ones That Mislead)

Most support operations measure what their platform reports by default, then manage toward those numbers. That is backwards, and it produces a specific pathology: teams that hit every target while customers get steadily less happy.

Here is what is worth managing, what each number actually tells you, and where the traps are.

Service level, stated properly

Service level is the percentage of calls answered within a target time, written as a pair: 80/20 means 80% of calls answered within 20 seconds.

It is the right headline number because it describes the experience of the distribution, not the average. But it has a well-known failure mode: when a team is measured only on service level, the rational behaviour under pressure is to answer quickly and deal with the problem badly. Answer, put the caller on hold, and the metric is satisfied.

Pair it with something that measures outcome, or it will be gamed — not dishonestly, but structurally.

Average speed of answer, with its distribution

ASA is the mean wait. Useful, but an average over a bimodal distribution is a lie: an ASA of 20 seconds is consistent with most calls answered instantly and a tail waiting four minutes.

Always look at the 90th or 95th percentile alongside it. The tail is where complaints come from, and it is invisible in the mean.

First contact resolution

FCR — the share of contacts resolved without a repeat — is the single most useful number in support, and the hardest to measure honestly.

It is useful because it correlates with satisfaction better than speed does, and because it is the metric that improves when you fix root causes rather than process more tickets.

It is hard because defining "resolved" and "repeat" requires judgement. A follow-up two weeks later about a different issue is not a repeat. A customer who gave up is not a resolution.

Typical healthy ranges for business support sit around 70–75%. Be suspicious of anything above 85%: it usually means the definition is generous rather than that the operation is exceptional.

Abandonment, split by wait time

Abandonment rate is the share of callers who hang up before being answered. Headline abandonment is less useful than its distribution:

• Callers abandoning in the first ten seconds are often misdials or people who changed their mind. Largely noise.
• Callers abandoning after a minute or more waited, decided you were not going to answer, and left. That is the number that matters.

Many platforms let you set a "short abandon" threshold and exclude those. Do it, and watch the long tail.

Average handle time: manage it, do not target it

AHT is the most optimised metric in the industry and the most misleading one to target.

Pushing AHT down produces rushed calls, more transfers, and more repeat contacts — which raises total cost while the metric improves. The cost moves somewhere the metric does not look.

AHT is useful as a diagnostic: a sudden rise suggests a new problem, a new product, or a training gap. It is useful for capacity planning. It is not a performance target for individuals.

If you want one number that captures efficiency honestly, use contacts per resolved issue — which punishes exactly the behaviour AHT targeting rewards.

Occupancy and shrinkage

Occupancy is the proportion of logged-in time agents are handling contacts. Below about 70% you are overstaffed; sustained above about 85% you are burning people out and your attrition will tell you so within two quarters.

Shrinkage — the share of paid time not available for contacts, covering training, breaks, meetings and absence — is the number most often underestimated in staffing models. Real shrinkage of 30–35% is normal. Plan with a number you have measured, not a number you hope for.

Customer satisfaction, with the sampling honesty

CSAT after a contact is useful if you accept that responders are not representative. Response rates are low and skew toward the extremes.

Treat the trend as the signal and the absolute value as approximate. A CSAT that moves from 4.3 to 4.0 is meaningful; the difference between 4.3 and 4.35 is not.

The ones to add that most teams do not have

Repeat contact rate by issue. Which problems generate multiple contacts? That list is your fix backlog, ranked by cost.

Transfer rate, and where transfers originate. High transfers from one entry point usually means a menu option is described misleadingly — a routing problem being absorbed by people.

Contacts per customer per month. Rising means your product or onboarding is generating work. Falling can mean either improvement or that customers have given up.

Longest wait, daily. One number, looked at every day. Averages hide the person who waited eleven minutes; the maximum does not.

How to use them together

Pick one headline per dimension and hold the rest as diagnostics:

• Accessibility: service level, with abandonment after the short threshold
• Effectiveness: FCR, with repeat contact rate by issue
• Experience: CSAT trend
• Sustainability: occupancy and attrition

Four numbers for the weekly review, everything else for investigation. A dashboard with thirty metrics does not get read, and the ones that do get read become the ones that get gamed.

The structural point

Every metric creates behaviour. Measure speed alone and you get speed at the cost of quality. Measure quality alone and you get thoroughness at the cost of waiting.

The only durable approach is a small set of paired metrics where improving one at the expense of the other is visible immediately — and a willingness to treat a worsening number as information about the system rather than about the people in it.