There are around 25 metrics service desk managers can focus on, and if the truth be known, totally confuse themselves with conflicting datasets. Jeff Rumburg and Eric Zbikowski in their whitepaper “The Seven Most Important Performance Indicators” exhort world-class service desk managers to realise when it comes to the analytics, less is indeed more when it comes to diagnosing trends in performance. Here I deal with what must be one of the hardest tasks a Service Desk Managers (SDM) will ever undertake: matching the staff supply to the call demand profile. This is not as easy as it first sounds. Also how well you solve this scheduling problem directly impact the key performance indicators “agent utilization” and “agent satisfaction” which in turn have the highest correlation with the majority of other key indicators. Finally agent turnover is one of the most costly things that a Service Desk can experience; and the shorter the scheduling horizon, the greater the agent discontent and turnover will be. So what is the problem exactly.
First you need to know the number of agents you are likely to need. This is almost exclusively worked out using the number of calls received during the day, often broken down into time periods e.g. 15, 30 or 60 minute intervals. There are then factors such as what is the average talk time per call; what is the average wrap/work time for each call; what is your absence rate; how long are lunches; how many hours do you operate; what is your occupancy target; and a few more. There are tools that will help you arrive at the number of staff while taking these factors into account KoolToolz CC-Modeller is one. When it comes down to it, even with this information you will only be able to obtain a simple Erlang result as an answer. As an example, if you had 35 calls per hour, with a 600 second average talk time, 30 second wrap-up time, and an 80/20 SVL you’d need approximately 10 people staffed and productive on the phones per hour. Rarely does the call profile remain constant for the day, typically there is a build up to a peak, a plateau, and then a decline. So the staff supply will probably be a variable requirement i.e. more and fewer staff at different times of the day.
So how do you work out the variable staff supply once you have identified the number of staff you need based on the number of incoming calls and other factors we have mentioned. Well you use shifts, periods of work, or more precisely overlapping shifts. There are only four types of shifts on this planet, and you can check out more information about them here. For now, all you need to understand is overlapping shifts will give you your variable staff supply; and CC-Modeller will also enable you to work out these overlapping shifts to deliver the number of people needed during specific time periods. For example here we have a profile of around 450 calls a day starting at 8am and finishing at 11pm when the operation closes down:
Using Erlang C calculation we can determine the staff supply needed each hour to deploy sufficient agents to handle the number of incoming calls. In this example, the overlapping shifts needed and the number of agents on each shift is as follows:
4 Agents working 8am-3pm (7 hour) shift
2 Agents working 9am-5pm (8 hour) shift
1 Agent working 10am-6pm (8 hour) shift
1 Agent working 11am-8pm (9 hour) shift
2 Agents working 1pm-9pm (8 hour) shift
3 Agents working 3pm-11pm (8 hour) shift
2 Agents working 5pm-11pm (6 hour) shift
1 Agent working 6pm-11pm (5 hour) shift
It is of course not that simple whereby we add up the number of staff on each shift, in this example a total of 16 Agents. We have to take into account days off, and because it is a 7 day week operation a rotation of shifts and days off is required to ensure an adequate staff supply. That means we need more than 16 agents, but how many more.
We will take a look at how to do this in Part 2 and create a schedule which will extend the agents schedule horizon to over 12 months, reduce short notice changes, enable longer term planning, and improve agents well-being and contentment to levels guaranteed to drive down those turnover figures to more acceptable levels – and while we are at it, look at a comparison of costs when we get it right; and when we get it wrong.