Looking in your rear view mirror isn't such a good idea while driving, but you do it when you forecast using only the history of a series. Most practitioner's dismiss causal modeling because it comes it's basis comes with the luggage needing many input series and most companies don't readily have all of these variables easily available for use. Causal modeling doesn't have to include 18 variables which we call "kitchen sink" modeling. If you can include just the important effects, the model Box-Jenkins stochastic features will capture the other inherent effects. More simply, if you can include just one variable that helps to explain the variability in what you are trying to predict then you have moved forward in your approach. 

Consider the following: Your marketing manager tells you that the sales of your soda product is effected by the following events:

• Day of the week 
• Price 
• Holidays
• Weather
• Other variables not identified by you 

How Autobox can help you and your marketing manager: 

•  Include a variable that adjusts the forecast for the day of the week
•  Detect a lead, contemporaneous, or lag effect for the price/holidays variable
•  Effect of weather on sales
•  Effect of Pulses, Seasonal Pulses, Level Shifts plus any other variables that may arise consistently

Remember that if you ignore the driving factors of your business then you could perceive these effects as outliers and unbeknownst to you, you could treat them as outliers since they seemed unusual. 

One of our clients who has embraced causal modeling found out that one store in pocket of America has high sales on Veteran's Day. This example is a store in a small town so Veteran's Day is still a important event and people respond to it. Our client does respond to this event by stocking up in anticipation of it and thereby acting upon the information learned in the causal process. Let's examine.