I am trying to maximize the transactions by optimally allocating the spend of different media vehicles. For this I am first building a Econometric model which would model the transactions by taking different media vehicles spend, economic variables and competitor spends as independent variables ( I am using PROC ARIMA and building a log-log model). Now in order to do the optimization I am using the results of this model in the PROC OPTMODEL. I am new to the optimization side and don't know much about OPTMODEL. I used the syntax given in the sas website and used OPTMODEL. I have a problem with the results that I am getting. The increase in the transactions in very very high. I am getting an ROI of more than 150%, which I think is quite not possible. For example, the transactions per week is around 50,000 but the OPTMODEL says that if the money is spend (among different media vehicles) according to what it says the transactions will go up to 125,000!! Any kind of help on this will be greatly appreciated.
Specifically:
My ARIMA model results has few insignificant variables....Do I use them in OPTMODEL? My ARIMA Model has few negative coefficients...Do I use them in OPTMODEL. If I don't use them, do I account for their spend in total budget when giving the constraints? I see that the huge increase in transactions is majorly driven by the intercept of the ARIMA Model. For example: If my model has an intercept of 10.64 the transactions that the OPTMODEL says I would be getting is 113,00 but if I decrease the value of intercept to 10, OPTMODEL says that the transactions would be 59,000 ( which is , according to me and the business, quite a believable number!!). Does this sudden decrease in transactions by a small change in intercept pointing towards a badly built model? Any help would be greatly appreciated!
Thanks in advance
Sohail Mohammad
