Demand Forecasting

Demand Forecasting

In this lesson, you’re expected to:
– understand how demand forecasting is used in operations
– learn about commonly used forecasting methods
– explore the demand forecasting process

Demand forecasting is the basis for operations planning decisions

Forecasting is a process to predict future events that forms the basis for several operations planning decisions, such as those related to production, inventory, human resources, and capacity. Its use is applicable to every type of business, such as retail, manufacturing, services etc.

Some examples of the use of demand forecasting:
– Governments need to forecast the number of children in a region order to build or expand schools.
– A book retailer needs to forecast the number of books sold so that they can plan their inventory levels.
– A management consulting firm has to forecast the number of projects in order to allocate or hire staff.

Some decisions only need short-term forecasts, such as deciding how many newspapers should be ordered the next day or next week. On the other hand, capacity planning, such as building a new plant, should consider long-term forecasts.

Mature products, such as salt or milk, are usually more predictable and easier to forecast; however, some products, such as fashion goods, are unpredictable and difficult to forecast.

However, we might be missing important factors using such an approach. An improved model (Figure B) could include additional information such as child mortality rate and migratory movements, which can be influenced by economic prosperity and other factors.

Forecasts are always wrong and the uncertainty level varies according to the level of aggregation and time frame

1) The time frame of the forecast

Long-term forecasts are usually less accurate than short-term forecasts, i.e. it is easier to predict how many units are likely to be sold within the next hour than within the next decade.

2) The level of aggregation

Aggregated forecasts are usually more accurate than disaggregated forecasts. This is because, in aggregated forecasts, the errors are sometimes canceled out.

As you can seen in the table below, the absolute total aggregated forecast error (2%) is less than the forecast error of any isolated product.

Overview of Forecasting Techniques

Enlarged version: http://bit.ly/2q6o4m7

Overview of Qualitative Methods

Enlarged version: http://bit.ly/2psbbqf

1) Time Series methods

Time series examines the pattern of past behavior of demand, taking its level, trend, and seasonality into consideration in order to forecast future demand.

Enlarged version: http://bit.ly/2oXircy

2) Causal methods

Causal methods evaluate complex cause-effect relationships between key variables in order to devise the demand forecast.

The example below is of a regression analysis showing the relationship between demand and the previous week’s temperature.
Source: Slack et al. (2010)

Enlarged version: http://bit.ly/2pWCzg6

Overview of a general process of demand forecasting