Profit From Your Forecasting Software

CHAPTER 4Curve Fitting and Exponential Smoothing

4.1 INTRODUCTION

When you have dedicated forecasting software, you won't have to set up formulae in spreadsheets like Microsoft Excel® or carry out calculations by hand. The software will do the hard work and produce your forecasts as requested. However, it's still worth understanding how these forecasts are obtained so you are aware of the rationale that underpins them. That way you'll have a better understanding of the strengths and limitations of different methods and know whether they are appropriate for your data.

In this chapter, we will look at methods that simply use your sales history to identify and project underlying patterns, such as trends or seasonal cycles, into the future. These methods – called time series methods – don't take any account of factors that might be driving sales, like advertising expenditure or price. Because they only process data on a single variable – past sales – they are referred to as univariate, or time series, methods. Despite ignoring potential drivers, these methods can be as robust and accurate as methods that are more complex. They also avoid the need to collect and record data on potential drivers, which in some circumstances can be expensive and time consuming. Better still, they can often be run automatically, which is a considerable advantage if you have hundreds or thousands of SKUs to forecast.

Here, we will look at two types of time series methods: those based on fitting curves ...