Call center KPIs: which metrics to measure and how to read them

What call center KPIs are and why a few are enough

Call center KPIs, or key performance indicators, are the numbers that sum up how your team is doing on the phone. Some look at the customer (how long they wait, how often they can't get through), some at the agent (how much time goes on conversations, how productive they are), and some at results (sales, appointments, problems solved).

The temptation is to measure everything. The usual result is a dashboard full of numbers nobody looks at. You're better off picking five or six indicators tied to what matters to you, understanding them properly and tracking them over time. A support desk will mostly watch waits and abandonment; a sales team will watch conversions and talk time.

One rule applies to everyone: never read a KPI on its own. A low handle time might mean efficient agents or customers being rushed off the line; high occupancy might mean a well-sized team or people at breaking point. Meaning comes from comparing different indicators and different periods.

Answer rate and missed calls

Answer rate tells you how many of the calls you received reached a person. It's the first number to check, because it measures the most basic thing: did the customer manage to talk to you?

Formula: answer rate = answered calls / received calls × 100. Missed calls are the other side of it: every call that went unanswered, for whatever reason.

The daily average nearly always hides the problem. Calls get missed at certain hours, on certain days, in certain queues. That's why it pays to look at answer rate hour by hour: 90% across the day can contain one morning hour where half the calls go unanswered.

Then separate the reasons. A call missed out of hours is different from one missed because the queue was full, or because the caller hung up after two minutes on hold. In Robottino's call reports, for instance, every call has a status (answered, no answer, out of hours, limit exceeded and others), and the inbound calls by hour chart shows where the losses cluster.

Queue abandonment rate

Abandonment rate measures how many people who entered a queue hung up before reaching an agent. It isn't the same as the missed call rate: it only covers callers who made it into the queue, and it tells you how much waiting costs you.

Formula: abandonment rate = calls abandoned in queue / calls entering the queue × 100.

Two adjustments make the number more honest. First, some people hang up within a few seconds, perhaps because they dialed the wrong number or heard the welcome message and changed their mind. Many call centers exclude these short abandons using a threshold of a few seconds; if you do, write it down and always use the same threshold. Second, keep abandons separate from planned exits, such as the maximum wait running out or a key to request a callback. Someone who presses 1 to be called back hasn't abandoned: they've taken another route.

Read abandonment alongside waiting time. If it climbs as waits climb, the issue is capacity; if it stays high even with short waits, look at what callers hear in the queue and whether the opening message is too long.

Average wait time and service level

Average wait time, often called ASA (average speed of answer), is how long answered callers wait on average. Formula: average wait = total wait time of answered calls / number of answered calls.

Averages mislead, though: ten calls answered instantly and one left in the queue for ten minutes give you a decent average and one furious customer. That's why service level is used as well, measuring the share of calls answered within a threshold you choose. Formula: service level = calls answered within X seconds / calls entering the queue × 100.

Some call centers use all queued calls as the denominator, others exclude short abandons: again, what matters is choosing a definition and sticking to it. It also helps to look at waits in bands, for example within 10 seconds, within 30, within a minute, within two and beyond. That's the view you get in Robottino's general report, and it shows straight away how long the tail of your worst cases is.

Where should the threshold sit? It depends on the kind of service and what your customers expect. A technical emergency line and a general information number have different needs. You might start, for example, by deciding you want to answer most calls within half a minute, measure for a few weeks and then raise or lower the bar.

Average handle time: talk, hold and after-call work

Average handle time, or AHT, measures how much of an agent's time a call takes on average, from the start of the conversation to the end of the work that follows it. Formula: average handle time = (talk time + hold time during the call + after-call work) / calls handled.

It's the key number for staffing: if you know how many calls arrive in an hour and how long each takes on average, you know how many agent hours you need. It's also one of the most misused KPIs. Pushing agents to cut it at any cost produces rushed calls, customers ringing back and problems left unsolved.

Use it instead to understand where time goes. A long conversation might point to a convoluted process worth simplifying, information that's hard to find or training that isn't finished. Long after-call work might mean too many fields to fill in. Comparing handle time between agents only makes sense if they take the same kind of calls.

In call queues you can set a wrap-up pause, a few seconds after each call when the agent receives no new queue calls so they can finish off: bear it in mind when reading times and occupancy.

Agent occupancy, availability and breaks

Occupancy tells you how much of their working time agents actually spend on calls. The most common definition looks at time when the agent is available or busy: occupancy = (talk time + after-call work) / (talk time + after-call work + available idle time) × 100.

There are variations, and you need to know which one you're reading. In Robottino's agent report, for instance, occupancy is talk time divided by time online, so breaks pull it down. Neither formula is wrong, but comparing numbers calculated differently leads to the wrong conclusions.

Low occupancy means agents spend a lot of time waiting for calls: perhaps there are too many of them in that time slot. Very high occupancy over long stretches means no breathing space between calls, with fatigue and mistakes on the way, and waits that blow up at the first peak. The aim isn't the maximum, it's a balance you decide while watching waits and abandonment too.

• Time online: from when the agent logs in to when they log out.
• Available time: online and not on a break, ready for calls.
• Break time: worth reading by type of break and by time of day.
• Talk time: actual time on the phone with customers.
• Schedule adherence: how closely real hours match planned shifts.

Conversion, first contact resolution and satisfaction

The KPIs so far describe how you work; these describe what you achieve. Conversion rate measures how many calls led to the result you care about: a sale, an appointment, a renewal. Formula: conversion rate = calls with a conversion outcome / calls handled × 100. To work it out, every call needs an outcome: with call outcomes you decide which ones count as conversions and read them by agent, queue or campaign.

First contact resolution (FCR) is the flagship support KPI: the share of requests solved without the customer having to call again. You can estimate it with a dedicated outcome chosen by the agent, by counting how many numbers call back within a few days about the same thing, or by asking the customer. Each method has limits, so pick one and use it consistently.

Customer satisfaction, finally, is measured by asking. A short post-call satisfaction survey with a keypad score gives you a number you can compare by agent and by queue. Read the average together with the number of responses: an average built on a handful of answers swings a lot from week to week.

KPIs for outbound campaigns

Outbound campaigns ask different questions. What matters isn't how long the customer waits, but how many contacts you reach, how long it takes and what you get out of it.

Contact rate is the share of leads you've spoken to at least once: leads with an answer / leads worked × 100. Average attempts per contact shows how hard you're pushing and what it costs to reach someone. Conversion rate by list shows which lists deserve your agents' time. SMS campaigns use similar indicators, with deliveries and clicks instead of answers: see the guide to SMS campaigns with tracked links.

If you use an automatic dialer, add the share of answered calls with no agent available: it's the price of pace and should stay low. How it changes across dialing modes is explained in the guide to predictive, progressive and manual dialers.

Summary table and how to set your own targets

The table sums up the main KPIs. It contains no industry target values, because they depend too heavily on the type of service: you build yours from your own data.

1. Measure without changing anything

   Collect two to four weeks of data to get a real baseline, hour by hour and day by day.

2. Pick a few indicators

   For support, say, answer rate, abandonment and service level; for sales, contact rate and conversion.

3. Set reachable targets

   Aim for an improvement on your own baseline, not a number you read somewhere else.

4. Review regularly

   Go through the KPIs weekly with the team and change one thing at a time, so you know what worked.

Frequently asked questions