The mismanagement of call center agents
'Average Handling Time' and 'Call Duration' metrics unfair to the performance quality
Callcenter agents are commonly being categorised by two metrics: the average duration or the average handling time of their call. Both metrics are, however, unfair to the true performance of the agent, since our research proves that the standard variation in call duration is huge. This variance is clearly caused by the question and the customer, and not by the agent.
A lot of call center agents are being categorised along 1 metric, the average duration of their call. Since even a child can understand that this is unfair, since 'afterwork' is most of the times a substantial part of the work of a call center agent, nowadays this metric has been broadened to average handling time, where handling time is the call time plus the afterwork.
Both metrics are, however, unfair to the true performance of the agent. Auditio has recorded about 300 conversations between customers and inbound call center agents. These conversations have been grouped to their nature, since a change of address for a subscription is inherently shorter than a change of address for a utility company. After this grouping Auditio ranked the conversations for their quality. This was being done using very detailed and tested forms where all aspects of the call can be rated. These aspects not only included items like 'tone of voice' and other soft skill items, but also included hard skills; was the answer correct? Was it an answer on a question that had been put forward by the customer?
Finally the team of Auditio zoomed in on those calls that received a perfect score for the hard skills as well as for the soft skills. The team then listed the duration of the call, where (from a customers' perspective correctly) time being put on hold, was measured as call time too.
The call handling time of 300 calls all being classified as 'handled perfectly':
(Y-axis: time in minutes)
The standard variance of these top calls (and hence top performing agents) was a whopping 104 on a (standardised) average of 150 seconds per call. This means that although only the best calls had been taken into account, the variation in duration is huge. Since all these calls where handled correctly, by agents that apparently do their work very well, it would be hugely unfair to classify them by the metric of Average Handling Time, since this variance is clearly caused by the question and the customer, and not by the agent.
This practice is, unfortunately, very common. The main reason is, we guess, because the metric is easy to lay your hands on if you are a call center manager. The telephony switch gives you they information about average handling time (almost) for free. The much fairer ranking, based on a precise form that clearly categorises the actions and answers of the agents as good or bad, takes a lot of time and an objective team of team leaders, both usually in short supply.