Simple Vs Weighted average accuracy

Many supply chain planners often get lost in accuracy calculations. The reason for struggling is that there are many ways to calculate accuracy. Picking the best option out of many is a real deal. Here are few examples ,which shows how different approaches yield different results.

Product	Actual Demand	Demand Plan	Difference	Accuracy
A	100	400
B	200	300
C	300	200
D	400	100
Total	1000	1000	0	100.00%

Accuracy :- 1- (ABS(Difference)/Actual Demand)

In the above example, it appears that a supply chain planner is doing really a great job because he is 100% accurate at demand planning. You tell me, do you really think he is a good planner ?  If you said "no", then you are correct.

The reason for accuracy to show 100 % is because he added all his products together and calculated accuracy on that. This is a very straight forward way of calculating accuracy, which most of the time tells incorrect result. (This is how you should calculate if you are a public company and you need to show it to your stakeholders :))

In order to find out the real accuracy at higher level it is always recommend to use weighted average rather than  simple average. In the above example, the planner was doing simple average. Now let's use the same set of data and calculate weighted average. you will notice the difference in accuracy at higher level.

Product	Actual Demand	Demand Plan	Difference	Accuracy
A	100	400	300
B	200	300	100
C	300	200	100
D	400	100	300
Total	1000	1000	800	20.00%

Accuracy :- 1- (ABS(Difference)/Actual Demand)

Damn ! Planner was only 20% accurate in planning. It appears he did it real bad !

Before going any further let's further analyze the difference between both calculations.  In our first example, planner simply added Actual Demand and Demand plan and then took difference. However, in the second example, planner calculated difference at the lower level, rolled them up to the higher level, and then, he calculated accuracy. The second option is best way of calculating accuracy.

I will let you know in my next blog why second option is the best option and how "Winsorizing" (?)  method helps in decision making for supply chain planner. I will also show you how to calculate Accuracy (Regular and Weighted) using SQL code in my next blog.