Forecasting

One of the critical business tool that have been growing in recent days with growth of the business intelligence through big data science is forecasting. According to a simple rule of statistical analysis is that all data are following the pattern of normal distribution, the need to analyze the old data to predict the short and mid-term future one. In this article we are going to have a brief overview on what is forecasting, the different types and what is the best method according to the data type.

Forecasting can be simply known as the process of predicting future data based on historical data. Businesses apply forecasting methodologies to decide how to distribute their budget lines for projected expenses for any forthcoming period of time. The term forecasting in supply chain usually correlated to the demand planning as a predictive analytics field.  This is typically start with what is the expected demand for the goods and services they offer; then proceeding with the quantities that should be produced to cover the market demand and offset any controllable variance in demand.

The central pointers to the forecasting process are uncertainty and risks; the uncertainty about the future and any unexpected events might incur the risk of supply chain and market sustainability like what happened in some countries during the beginning of COVID-19 epidemic when market was shorted in medical supplies, like face masks and alcohol disinfectant sprays, due to high unpredictable consumption.

The major types of forecasting techniques are

·         Qualitative forecasting which is concerned about limited scope forecasts as they are very much dependent on surveys and opinions of experts.

·         Quantitative forecasting which is concerned about quantitative data like quantities sold or dollar sales. 

The use of quantitative forecasting method is the common in the world of data. There is multiple common quantitative methodologies to perform forecasting such as moving averages, exponential smoothing, and trend projection. The selection of a method depends on several factors like the framework of the forecast, the wanted degree of accuracy, the accessibility of historical data, and the forecast’s time horizon.

The quantitative forecasting methods can be grouped into the moving averages, exponential smoothing, and trend projections. The difference between moving averages and exponential smoothing is that the moving average is giving an equal weight to all observations in the study. While in exponential smoothing, the weight is assigned in a decreasing way over time. The trend projection is used when there is obvious style in the data over a specific time period either linear or exponential.

As a demand planner, the method selection is depending on the behavior of the data. For example, for a monthly sales of a specific product, does it sold every month with high coefficient of relation to the average; then the moving average method seems to be suitable there. The same is for seasonal products, to compare current season with the previous season using the same time frame, i.e. weeks, months, or quarter. On the other hand, the data with low coefficient of relation to the average; we can use trend projection methods such as linear regression, or exponential regression methods.