Calculating a Service Level Goal

Service level goals are key performance indicators in most support and call centers. These goals measure the time callers spend in queue. In benchmarking surveys, the most common goal given by respondents is to answer 80 percent of calls within 20 seconds. Another common response is an average speed of answer — or ASA — of 30 seconds or less. This article is focused on ensuring a clear understanding of the calculations that make up these statistics and the varying results that different statistical methods deliver.

Service level measurement is a science. The typical automatic call distribution, or ACD, system offers two measures of queue time for callers. The first is service level — factor expressed as X percent of calls answered within Y seconds. So, if the center met the goal of 80 percent answered within 20 seconds, that means that 20 percent of callers waited at least 21 seconds. You know how many were queued longer than 20 seconds but not how long these callers actually waited.

The other queue time goal is ASA. This goal is expressed as an average number of seconds, and the most common goals cited are less than 30 seconds. ASA is the average queue time for all callers, whether or not they queue. So, if 60 percent of 100 callers are answered immediately by idle agents, that is 60 times zero seconds in the calculation.

Assume the other 40 percent of callers who do queue wait between one and 300 seconds with an average of 60 seconds. That is a total time in queue of 2,400 seconds (40 callers X 60 seconds). This 2,400-second total is then divided by the 100 total callers, and the resulting ASA is 24 seconds. ASA does not really tell you how many queued or how long they waited, just that the overall average was X seconds.

Service level measurement is an art. To assess the speed of answer, two more factors must be considered — the way the calculation is done and the interval over which the measurement is calculated.

When does the ACD system begin measuring the queue for calculating the service level or ASA, and which calls are included? It differs from one ACD to another. For example, the queue time can begin to count when the call enters the ACD for a specific agent group and the caller hears ringing. In another ACD the ringing could be ignored, but the recorded announcement will be included in the service level or ASA calculation. The ACD could also be set to ignore the recorded announcement and start counting when it ends. And, finally, the ACD could be set to ignore a standard interval (such as five seconds), then start the count.

To make it even more interesting, on the service level goal, the system can be set to measure only those calls that are actually answered within 20 seconds or can include all calls, including those that are abandoned within the 20 seconds, as well. Do you know what your system is set for and what your options are?

The second factor is the interval. Centers absolutely committed to their speed-of-answer goals measure each hour or half-hour and determine what percentage of these intervals meet the goal. In many centers, the average is calculated across a whole day so that if the peak hours are awful and the low hours are great, it all averages out. Still more centers average across a whole week so the peak day's bad service is buried in the slower days' great results. In some extreme cases, it is even possible that a center averages across a month or even the whole year, rendering any measurement meaningless.

The longer the averaging interval, the more that dreadful service can be hidden. Of course, these busy intervals are when the most customers are affected. When the interval is long, it enables centers that give terrible service early in the interval to average out to acceptable levels by overstaffing later in the period to give more than fabulous service. This is neither good service nor cost-effective.

It is important to understand how your call center measures service levels to ensure those measurements give you the information that truly reflects performance. The time to do it is now because performance management will only get more complicated.

As you move to include other kinds of contacts in your centers, such as e-mails and Web chats, you need to consider how you will measure the service level.

Some will need similar queue measures, while others will be driven by time to complete response. But knowing what to calculate and how the systems work will be key to setting your goals and meeting your customers' expectations.

Related Posts