Choosing a Predictive Model for Call Center Forecasting

Scheduling is one of the most critical functions of a call center manager. In order to meet expected customer demand and call volumes, the contact center needs to know how many agents, and with what skills, to schedule. The best way to figure this out is to do a bit of guessing: not random guessing, but educated guessing based on historical data, or volumes from past weeks, months and years. (If your contact center is too new to have any historical data, you may need to find someone’s else’s data to work with.)

How is Forecasting Accomplished?

There are several different ways to forecast based on which predictive model you choose, according to the Society of Workforce Planning Professionals. The four different models are described below:

The point estimate. This simple model assumes that any point in the future will match the corresponding point in the past: “The week before Christmas this year will be the same as the week before Christmas last year.” While it’s simple and straightforward, obviously it doesn’t accommodate the fact that there could be upward or downward trends in calling patterns, or the fact that the same period last year could have been atypical. If you have many years of data to use, this might be a good option for you.

Averaging. Mathematical averaging can range from a simple average of several past numbers, to a moving average where older data is dropped out when new numbers are available, according to the SWPP. The most accurate averaging approach involves weighted averaging, where more recent events are given more weight or significance than older events. In other words, if you’re averaging the last eight years, you may want to use the more recent years more heavily by giving them more weight (say, 80 percent) in the average, and assign more distant years with lower weights.

Time series. The most complex and possibly the most accurate method is called “time series analysis.”  This method takes historical information and applies information about trends (market sentiment, product popularity, marketing, etc.) as well as seasonal or monthly differences. Most large contact centers employ this method, as do many companies that build call center scheduling solutions. The idea is that call volume can be influenced by a variety of factors over time (trends) and that each of the factors can be isolated and used to predict the future, though it’s not easy to isolate the effects of trends. This is something an automated call center scheduling solution can help with.

Call center scheduling is a lot of science and a little bit of art. (And, sometimes, some magic.) A good call center scheduling solution can use the best of each of these models to build the best possible forecast so you’re not caught understaffed or overstaffed. For more information, download Verint (News - Alert) Monet’s white paper, “Contact Center Forecasting and Scheduling Best Practices.”

Edited by Maurice Nagle