How to Calculate Service Level in a Call Center

Here is how I calculate service level in a call center. In short: service level is the share of calls or chats answered within your target time. The formula I use is contacts answered within the threshold ÷ (contacts offered − contacts abandoned before the threshold) × 100. 3,150 answered within 20 seconds out of 4,000 offered, with 120 early hang-ups, is 81.2%. The daily number can look fine while customers wait for minutes every morning, so I always read it by half-hour interval.

When I was head of customer support, service level was the number on the wallboard above the team. It was also the number that fooled me the most. This is how I calculate it, the three versions of the formula you will meet, and the month our "81%" hid a morning where almost half the callers waited too long.

How I calculate service level

Service level = answered within the threshold ÷ (offered − abandoned before the threshold) × 100

You need three numbers from your phone or chat system for the period:

• Contacts offered: every call or chat that entered the queue.
• Answered within the threshold: answered by an agent within your target time, for example 20 seconds.
• Abandoned before the threshold: customers who hung up or left before that time passed.

Example for one day: 4,000 calls offered, 3,150 answered within 20 seconds, 120 abandoned within 20 seconds. Service level = 3,150 ÷ (4,000 − 120) × 100 = 81.2%.

An 80/20 target means 80% of calls answered within 20 seconds. It is the classic voice target. Chat teams often use 80% within 60 seconds. Email is not measured this way; I track it as first response time instead.

The service level calculator does this from your three numbers and has a free Excel version.

Service level today

The daily number and the formula

Clarivoxx analysis

Calls offered4,000

Answered within 20 s3,150

Abandoned within 20 s120

Service level81.2%

The formula

Service level = answered within 20 s ÷ (offered − abandoned within 20 s) × 100 = 3,150 ÷ (4,000 − 120) × 100 = 81.2%.

Three formulas, three different answers

The formula your system uses changes the result, so check it before you compare teams or vendors. Same day, same calls (220 abandons in total, 3,780 answered):

1. Answered within 20 s ÷ offered: 3,150 ÷ 4,000 = 78.8%. Every abandon counts against you, even a caller who hung up after two seconds.
2. Answered within 20 s ÷ (offered − early abandons): 3,150 ÷ 3,880 = 81.2%. Early hang-ups are left out; late ones still count. This is the one I use.
3. Answered within 20 s ÷ answered: 3,150 ÷ 3,780 = 83.3%. Every abandon disappears. It flatters the team, so I never report it alone.

Whatever you pick, write the rule down and keep it. A 2-point "improvement" is often just a formula change.

The month our 81% was hiding a problem

One month our service level was 81.2% for the day and the wallboard was green. Complaints about waiting, though, were the second most common comment in our customer survey. The split by time of day explained it.

Service level by time of day

The morning was far below target

Clarivoxx analysis

0%50%100% of best window

08:00–10:00
52.6%

10:00–12:00
82.5%

12:00–15:00
98.9%

15:00–18:00
98.8%

The first two hours carried 30% of the calls at 52.6%; the afternoon ran at almost 99%Window service level = answered within 20 s ÷ (offered − abandoned within 20 s) in that window × 100

08:00–10:00: 600 ÷ (1,200 − 60) = 52.6%. Erlang C for 600 calls an hour at 4 minutes handle time: about 46 agents for 80/20; about 43 were on the phones.

Between 08:00 and 10:00 we received 1,200 calls, 30% of the day, and only 52.6% were answered within 20 seconds. In the afternoon the same team answered almost 99% in time, because we had more agents than calls. The afternoon was so good that it pulled the daily average above target.

I put the morning numbers into the Erlang calculator: 600 calls an hour with a 4-minute handle time needs about 46 agents on the phones for 80/20. We had about 43 in that window, because most shifts started at 09:00. Three agents short does not sound like much, but at that load it is the difference between about 84% and under 60%.

Why the daily number lies

Service level is not an average you can add up. Two things make the daily figure misleading:

• Busy intervals carry more customers. A bad hour at 600 calls hurts more people than a perfect hour at 280 calls helps.
• Over-staffed hours hide under-staffed ones. 99% in the afternoon and 53% in the morning blend into a calm 81%.

So I look at service level for every half-hour, and I count how many intervals fall below 70%. That count tells me more than the daily figure.

I also read it next to two other numbers: the abandonment rate, because customers who give up do not show in the "answered" figure, and average speed of answer, because a few very long waits can hide behind a good percentage.

What I changed

Fix plan

Same hours, better placed

Clarivoxx analysis

Most shifts started at 09:00, after the peak beganFour shifts moved to 08:00–17:00Workforce plannerOwner assigned

Breaks overlapping in the busy hoursNo breaks before 10:30, max 15% of the team off at onceTeam leadsOwner assigned

Training and meetings in the morningMoved to 15:00–17:00Head of supportOwner assigned

Bad intervals noticed too lateAlert when a half-hour drops below 70%Support operationsOwner assigned

We did not hire anyone. We moved the hours we already had:

1. Moved four agents' shifts one hour earlier, from 09:00–18:00 to 08:00–17:00. The late afternoon had spare capacity anyway.
2. Staggered breaks and lunch so no more than 15% of the team was off the phones at once, and none before 10:30.
3. Moved training and team meetings out of the morning peak into the 15:00–17:00 window.
4. Set an alert when any half-hour interval dropped below 70%, so the team lead could pull agents from email back to the phones.

My service level dashboard

I update this table every Monday for the previous week. "This month" is the first full month after the shift change; "last month" is the month of the problem.

Metric	Target	This month	Last month	Change
Service level, whole day	80%	87.1%	81.2%	+5.9 pts
Service level, 08:00–10:00	80%	81.5%	52.6%	+28.9 pts
Half-hours below 70% (per day)	0	0	4 of 20	−4
Abandonment rate	under 5%	3.1%	5.5%	−2.4 pts
Average speed of answer	30 s	19 s	41 s	−22 s

The afternoon dropped from about 99% to 90%, which is still above target and exactly the trade we wanted.

What to do this week

Export last week's calls by half-hour: offered, answered within your threshold and abandoned before it. Calculate service level for each interval, count how many are below 70%, and look at what time they happen. If they cluster at the start of the day, the fix is usually shift start times, not more people.