Having the right forecasting strategy is probably the most important thing to consider in practical forecasting. The strategy that is right for you might not be right for anyone else, and a forecasting strategy for sales of an individual product will probably not be appropriate for forecasting overhead expenses. This page will provide an overall checklist of things to consider as you develop the strategy that is right for you and your application.
What Do You Need And Why?
First, it is very helpful to write down what you need and why you need it. This subject is expanded in the forecast requirements page, and is an important first step.
Select The Right Structure
Second, with the requirements in mind, develop a structure that meets your needs and is as simple as possible. Start at the highest level and add only as much complexity as is needed. For example, if you're concerned about a revenue forecast, can you meet your needs with just a trend model? If so, stop with that. If not, does it have obvious seasonality issues that can be adjusted out sufficiently with a simple seasonality factor? If so, stop there. A forecast is only as strong as its weakest link (the assumptions), and every piece that you add to the model is another link. So don't add any more links than you need. And then go validate every one of them.
Third, build the forecast and validate it with historical data (link). By the way, this might be a good time to revisit the point that all forecasts are wrong. If the accuracy does not appear to be within your business requirements, is it feasible to change the business sufficiently to handle this level of accuracy?
Document It All
Fourth, after building the forecast, document the strategy and the structure, and save it for the next time you need to remember what you did and why you did it.
Management Is Key
Fifth, establish your management plan (link). All forecasts must be tracked on a regular basis, and this might be the most important step of all. By looking at the variance each period, you develop a solid feel for how accurate the forecast really is. As you use the forecast to make decisions, recognize that the actual value could easily be anywhere in that range.
Return to Forecasting Strategy Overview