I have a couple of item demands, and I am trying to use THREE different forecasting methods on each of them. Then, I calculate MAPE, MAD, and MSE for each method -- a total of NINE measurements per item. The purpose of this is to choose the best performing technique for each item.
The problem is that there is high variability in the demand and the methods fail to do their job. I want to know how useable the methods actually are.
For example, I have a MAD of 16 on one of the forecasts, but the average demand for this item is 17. Am I looking at MAD in "correct" context (AVG demand)? If yes, this is horrible result, right?
Any additional information I can provide.