The Monthly Forecasting Meeting: - Measuring Forecast Accuracy
The topic is forecast accuracy. There are many ways we can measure this and many discussions focus on the maths rather than the insight.
By on Apr 28, 2009The topic is forecast accuracy. There are many ways we can measure this and many discussions focus on the maths rather than the insight. Here's what happened when Fred the Forecaster, Colin the Controller and Pete from Production met recently...
The Forecasting Meeting
Number 17. Reporting Forecast Accuracy
Fred: OK guys.good to see you all today. There is only one item on the agenda how to report forecast accuracy.
Colin: Yup. This is important we need to get clear messages across to management on how good our forecasts are.
Pete: I agree, we need to know where to focus our attention across the product range.
Fred: OK. Now this is usually where I get the flipchart out and write down some equations that we all did in high school but have now more or less forgotten.
Colin: Theres another way?????
Fred: Well this is a bit revolutionary and spooky but I thought we might kick off by deciding what it was we were after. What do we want the number to tell us?
Colin and Pete: ...How accurate the forecasts are.!!!!
Fred: Well, thats progress. So we are not worried about sales accuracy?
Silence.
Fred: So all the questions we need to answer are about how good (or bad) the forecast was.????
Colin and Pete: Yesss,,,,
Fred: OK then. The question the numbers have to answer is. Was the forecast above or below the actuals?
You will note I have not included equal to to maintain realism. If the answer is above we need to number to be positive. If the answer is lower the number needs to be negative.
Were making progress here.
Colin: Wellthat makes sense if we over forecast its a positive number if we under-forecast its negative.
Fred: Spot onand there is more. The next question is how do we express this difference quantitatively?
Pete: Well I want to know by how much the forecast was wrong as we have mentioned before the sales are what they are they can never be wrong.
Fred: OK try this. If we get a positive number , x%, lets say this tells us that You have over forecast by x percent of your original forecast.'
Pete: And, alternatively x% You have under forecast by x% of the original forecast.'
Colin: Yes I see. Lets do this with numbers. You know I like numbers
Fred: Two situations. One the forecast is 120 and the sales come in as 100. Two the forecast is 120 and the sales come in at 150
The first index is positive. The over forecast is 20 and that is 17% of the original forecast. +17%
The second index is negative. The under forecast is 30 and that is 25% of the original forecast. -25%
How does that feel.
Pete: It makes sense to me the focus is on how wrong was the forecast and the signs are intuitively in the right direction.
Colin: I agree.
Fred: OK and now its simple to choose the maths to match the requirements. Rather than the other way around..
Forecast Error (%) = (F-A)/F x 100
A = Actuals F = Forecast
See you next week guys