7 Replies Latest reply: Sep 30, 2016 12:05 PM by Sangeetha Shanmugham RSS

    Weighted Accuracy rollup

    Sangeetha Shanmugham

      Hi All,

       

      I have been stumped on this weighted forecast accuracy for few days. Any help would be appreciated.

       

      Let's say we have the below hierarchy.

      Company->Division->Group->Model->Category->Item Group->Fcst Product Classification->Item.

       

      We have two types of accuracy - Normal and weighted accuracy. Normal accuracy is where all items have the same weight (1) and weighted accuracy is where items under Fcst Product Classification have different weights.

      Fcst Product Classification

      Weight

      High Revenue

      0.35

      Commodity

      0.25

      EOSPending

      0.10

      Legacy

      0.07

      Build to Order

      0.08

      High Growth Potential

      0.15

      If the items do not fall under these category, they would have weight value 1.

       

      I'm trying to calculate the weighted accuracy using the formula from Fcst Product Classification level.

                     1-∑ (error*weight)/demand, where error=sum(Forecast)-sum(Demand)

       

      1-sum(aggr((sum(aggr(fabs(sum(FORECAST-QTY)-sum({<Flag_Type=>}Demand)))/sum({<Flag_Type>}Demand)) * Forecast_weight,Item),[Fcst Product Classification])).

       

      The weighted accuracy works fine on Fcst Product Classification level, but when I try to roll up the accuracy to higher hierarchy, it changes. I want the accuracy to be consistent on all levels. How do I calculate the weighted accuracy and roll up the percentage to higher levels?

       

      Thanks,

      Sangeetha