I hope this is the right category for posting my query.
I was looking for some technical advice or input on an issue we were facing while implementing forecasting solution using APO DP statistical models.
Problem:- Using seasonal or seasonal trend models , the Alpha. Beta , Gamma parameters are currently set to as low as 0.2-0.18-0.5 as Alpha Beta Gamma. With these parameters the forecast generated looks good. This was considering 3 years history and i am predicting 2 years of Forecast. In the next month i re-run stat forecast using same parameters, and now my forecast results show a very disproportionate increase in trend in coming year by almost 60% increase . The recent month that got added was the highest sales month for this year and has been the same in past year on year. But the forecast that is getting generated post adding this recent month history showing a 60% increase looks way too off.
I tried to adjust the parameters now and had to bring it down to 0.2-0.05-0.5 Alpha-Beta-Gamma and the forecast now looks good again.
How can we quantify the impact these parameters are having on just a single months added history such that the end users know how to handle their products in the future incase such variations in history do pop up?