Service level calculator

How this is calculated

Service level is calls answered within the target time divided by calls offered. The common alternative divides by calls answered instead, which excludes abandons — and abandons are the callers who waited longest, so it systematically flatters the result.

serviceLevel        = answeredWithinTarget / offered      ← strict
serviceLevel(excl.) = answeredWithinTarget / answered      ← flattering
abandonRate         = abandoned / offered
ASA                 = totalWaitSeconds / answered

Worked example

1,000 calls offered, 940 answered, 60 abandoned, 780 answered inside 20 seconds. Against calls offered that is 78% — a 78/20 service level. Against calls answered it is 83%. Same period, same data, five points apart.

Which convention your ACD uses matters

Excluding abandoned calls removes exactly the callers who waited longest — the group the metric exists to protect. It is not dishonest, and plenty of platforms default to it, but it does mean two centres quoting the same number can be delivering visibly different experiences. Check which one yours reports before comparing against anybody else's figure, or against a target you inherited.

Service level beats average speed of answer because it captures the distribution rather than the mean. A queue where most callers are answered instantly and a handful wait four minutes can post the same ASA as one where everybody waits forty seconds — very different experiences, identical average. And 80/20 itself is a convention that spread by repetition, not a researched optimum; pick the target your customers and budget actually justify.

Frequently asked questions

How do you calculate service level?

Divide calls answered within your target time by calls offered. If 780 of 1,000 calls offered were answered within 20 seconds, service level is 78% — usually written 78/20. Whether abandoned calls belong in the denominator is the part people disagree about, and it changes the answer materially.

Should abandoned calls count against service level?

Under the strict reading, yes — they were offered and not answered in time. Dividing by answered calls instead removes exactly the callers who waited longest, which is the group the metric exists to protect. On the same data that convention can add five points or more. Neither is wrong, but comparing across the two is meaningless.

What does 80/20 mean?

80% of calls answered within 20 seconds. It is the most widely used target in the industry and a convention rather than a researched optimum — Call Centre Helper says so explicitly. Plenty of centres run 90/15, and others deliberately accept less where cost matters more than speed to answer.

Why measure service level instead of average speed of answer?

ASA is an average and hides the shape of the wait. A queue where most callers are answered instantly and a handful wait four minutes can post the same ASA as one where everyone waits forty seconds. Service level captures the distribution, which is what callers actually experience.

Related reading