“Service level 80/20” is probably the most familiar phrase in any voice process. It means 80% of calls answered within 20 seconds. What it does not tell you is which calls are counted in the 80%, and that detail can move the reported figure by several percentage points without a single customer having a different experience. Abandon rate has the same problem. This article walks through one day of data and computes the numbers in the ways contracts and dashboards actually do.
The raw data for one day
A domestic banking helpline queue records the following for a single day:
- Calls offered (entered the queue): 1,200
- Calls answered by an agent: 1,080
- Calls abandoned (caller hung up before answer): 120
- Of those abandons, 30 hung up within 5 seconds of entering the queue
- Of all abandons, 50 hung up within 20 seconds (this includes the 30 above)
- Calls answered within 20 seconds: 860
“Offered” here means the call reached the agent queue. Calls that ended inside the IVR menu, or that were fully resolved by self-service, are usually not counted as offered to agents. That is the first definitional choice, and it is worth confirming before anything else.
Abandon rate, two ways
The plain version is abandoned divided by offered: 120 ÷ 1,200 = 10.0%.
Many contracts exclude short abandons, calls that drop within a few seconds. The reasoning is fair: a caller who hangs up after three seconds probably dialled by mistake or got distracted, and no staffing level would have helped. Removing the 30 short abandons from both the numerator and the denominator gives 90 ÷ 1,170 = 7.7%.
The threshold for a short abandon varies. Five seconds is common; some agreements use ten or more. The longer the threshold, the more genuine frustration gets excluded, so a generous threshold is something a client should look at carefully.
Service level, four ways
| Formula | Calculation | Result |
|---|---|---|
| Answered within 20s ÷ offered | 860 ÷ 1,200 | 71.7% |
| Answered within 20s ÷ (offered − short abandons) | 860 ÷ 1,170 | 73.5% |
| (Answered within 20s + abandoned within 20s) ÷ offered | 910 ÷ 1,200 | 75.8% |
| Answered within 20s ÷ answered | 860 ÷ 1,080 | 79.6% |
Same calls, same customers, same agents. The reported service level ranges from 71.7% to 79.6%. Against an 80% target, the first looks like a serious miss and the last looks like a near-hit.
The last formula deserves a specific warning. Dividing by answered calls ignores abandons entirely, so a queue where hundreds of callers give up can still report a healthy service level. In the extreme, if every caller who waited more than 20 seconds hung up, this formula would show 100%. It is not a dishonest formula when used alongside abandon rate, but on its own it rewards losing customers.
The third formula treats a call abandoned within the threshold as “met”, on the reasoning that the centre was not given the chance to answer within target. Some workforce teams prefer it; others think it flatters the result. What matters is that both sides of the contract use the same one, and that it is written into the SLA schedule rather than left to whoever builds the report.
Averages hide the worst waits
Average speed of answer (ASA) is often reported next to service level. It is the total wait time of answered calls divided by the number of answered calls. The problem is that waiting times are heavily skewed. Take ten answered calls with waits of 2, 3, 4, 5, 5, 6, 8, 10, 12 and 145 seconds. The ASA is 20 seconds, which sounds reasonable. Nine of ten callers waited 12 seconds or less, and one caller waited nearly two and a half minutes. Service level at 20 seconds is 90% here, which describes the typical experience better, but neither number tells you about the person who waited 145 seconds.
That is why mature reporting adds longest wait or a high percentile of wait time per interval. A queue can hit 80/20 for the day while having one 45-minute interval where callers waited six minutes, which is exactly the interval that generates complaints and social media posts.
Daily, interval and monthly: the aggregation question
Service level is usually managed in 15- or 30-minute intervals but reported to clients daily or monthly. Aggregation changes the picture. A monthly figure weights every call equally, so a terrible Monday morning can be averaged away by quiet Thursday evenings. Some contracts therefore specify that the SLA must be met on a certain percentage of days or intervals, not just on the monthly total. Others require it per interval during peak hours only.
The method of aggregation also matters. The correct monthly service level is total calls answered within threshold for the month divided by total offered (or whatever denominator applies). Averaging 30 daily percentages gives a different result whenever daily volumes differ, because a 200-call day counts as much as a 2,000-call day. It is a small error that shows up surprisingly often in hand-built Excel MIS.
How the two metrics move together
Service level and abandon rate are linked through customer patience. When queues build, waits rise, more callers give up, and abandon rate climbs. But the relationship is not fixed. Customers calling about a blocked debit card will wait far longer than customers calling to check an order status. A process with patient callers can show a poor service level and a low abandon rate at the same time; one with impatient callers can show the reverse.
Abandons also generate repeat calls. A customer who hangs up after a minute usually calls back, so a high abandon rate inflates the offered volume for the next few intervals. That makes the next hour harder to staff and can push service level down further, which is why a bad morning on a voice process often stays bad until well past the peak.
When reading any service level figure, three questions sort out most confusion: what is the denominator, are short abandons excluded and at what threshold, and over what period has it been aggregated. With those answered, two dashboards that seemed to disagree usually turn out to be describing the same day.
Written by the RP Strativa Team. Figures in the examples are illustrative.
Leave a Reply