Causal Modeling

Causal Modeling is the use of independent explanatory variables to predict your demand. Software packages also refer to this as an econometric modeling or advanced modeling or structural models. Most forecasting and demand planning software rely on simple time series models that leverage the past demand observations to forecast the future demand. This is a time-tested proven method.

However, there may be external factors that drive the demand in a systematic fashion. If your business experience indicates that there are indeed factors that drive demand, then you need to explore the data availability and predictability of these external factors.

The most common mistake is to look for a very obvious external variable that explains your business demand and potential such as the GDP, or interest rates or even the price of oil. Although the historical data for these entities are easily available, it is impossible to forecast them. You can obtain external expert forecasts but then you are complicating a simple forecast problem and perhaps made it less forecastable.

The best situations to use causal modeling can be some of the following:

1. You expect the average price of your product to go up in Q2 and Q4 of next year. Although you made a forecast about price, it is still a policy variable and under control. Perhaps very accurately forecastable
2. The company always promotes the products at the end of the quarter.

   To understand structural modeling or multiple regressions and Discrete variable modeling, the following sources will be helpful

• Manual for univariate and multi-variate statistics
• Multiple Linear Regression from Statsoft
• Discrete Variable Regression Models
• An interesting theoretical approach to Linear Regression Models