I am trying to forecast future web sales for my company. A subset of that is "keycoded" revenue that can be traced back to particular advertising campaigns. The marketing dept. produces forecasts for keycoded revenue, and I would like to use those as an explanatory variable to predict overall web sales.
My concern is this: there is almost a perfect relationship between *actual* keycoded revenue and *actual* overall web sales. Obviously, I won't have the benefit of knowing future actuals when making forecasts; I will have *forecasted* keycoded revenue instead.
So my question is, would you prefer to use historical actuals or historical forecasts of keycoded revenue when developing the forecasting model?
I'm not completely sure I'm getting your point, but here are some thoughts anyway:
when using explanatory variables in your models it will be crucial to have future values of these variables in order to calculate the forecast for the dependent variable.
As such I think I would go with actual numbers, as they reflect what was happening in fact - not the "wishful thinking" of ours.
What you will need to do is to create forecasts for your independent variable first (if you don't have future estimates - or use the forecasts of your marketing folks as future values) and then create the prediction of the dependent variable.
In fact, you might consider doing a what-if analysis, i.e. testing what happens if you change the future values of your dependent variables.
Note that this is an out-of-the-box feature of SAS Forecast Studio - which is part of SAS Forecast Server.
Registration is open! SAS is returning to Vegas for an AI and analytics experience like no other! Whether you're an executive, manager, end user or SAS partner, SAS Innovate is designed for everyone on your team. Register for just $495 by 12/31/2023.
If you are interested in speaking, there is still time to submit a session idea. More details are posted on the website.