This blog is the third in a three-part series focused on forecasting in the contact center. The first installment discusses why forecasting is both an art and a science; the second provides a useful acronym to ensure more accurate forecasting. Executives and workforce managers alike consider some kind of forecast in virtually every decision they make.
Numerous forecasting techniques help leaders handle the challenges that crop up in an increasingly complex contact center.
Among them are the weighted moving average model, which is usually well-suited for stable historical data; the Box-Jenkins AutoRegressive Integrated Moving Average (ARIMA) model, which is often used to handle complex time series forecasting situations in which the basic pattern isn’t clearly apparent; and exponential smoothing, which can be particularly useful when historical data has stationary patterns and when you can assume that the past will continue to influence and represent the future. Modern forecasting systems can handle incredibly complex demands, weighing and incorporating any range of source, input and historical and real-time data.
Some of the most common applications of forecasting involve long-term needs: staff planning, real estate planning and budgeting, for example. Others, however, take place over a shorter time frame: in the contact center, short-term forecasts largely center on scheduling. Here’s what you need to know:
- Long-term forecasts are typically weekly or monthly projections made out over the next year to 18 months or between three and five years. They’re maintained and updated as new data arrives, usually weekly or monthly; when conditions are highly variable or unknown, updates occur more often – weekly, rather than monthly.
- Short term forecasts, as mentioned earlier, are often used for scheduling; schedules are built based on forecasted volume. Short-term forecasts typically cover one to 12 weeks and are reforecast weekly as new data comes in – even if you have to lock in a forecast four, eight or even 12 weeks out, you should be updating them as you go to be better prepared for what is likely to happen. If you find that your forecasting isn’t accurate – or if you receive news that will alter your operation, such as a change to your business – be prepared to make daily adjustments to the forecast for the remainder of the week and the following one.
During periods in which you have a lot of unknowns, you can increase the accuracy of your forecasting by producing shorter forecasts (e.g., 12 one-week forecasts instead of one 12-week forecast) or potentially generating single-day forecasts for remainder of the week.
You can also make manual adjustments to remaining forecasts when conditions change. One technique is to let the trend stabilize (if it happens once, it’s an anomaly; two times, an interesting occurrence; and three times, a trend for which the forecast needs to be adjusted). You then split the difference: If the trend over three periods is over by 5%, adjust the forecast by 2.5% and watch for another period or two. This helps you avoid making “whiplash” adjustments back and forth.
As you can see, forecasting isn't an activity that is done once and then set aside. It’s a dynamic process with frequent updates – and one that’s critically important to the customer experience and your organization’s bottom line.
Learn more about how NiCE WFM puts AI and machine learning to work in your contact center to understand and plan for future workforce needs efficiently and accurately.
