Measuring Forecast Accuracy

How should you measure forecast accuracy, for example. A salesman will measure his accuracy based on what he believes he will sell, he will change his forecast instantly if the situation changes, therefore he measures against a live forecast. A Buyer or Inventory Controller will measure an accuracy based on what the forecast was at the point where lead times triggered an order, perhaps 8 weeks earlier....a live forecast may have changed. A planner/scheduler would be somewhere in between, compiling a production schedule against what the forecast showed form anywhere form a week to 6wks before the product is sold.....any opinions?

dear John,

WIll you please tell us what reorder point formula you are using in your operations to control stock replenishment process?

thanks,
jon



quote=John Hardwick, Tuesday, 15 Jul 2008 20:47]You are right, measuring forecast accuracy should depend on what you are using the forecast for.

In our business (pharmaceuticals) we measure the actual demand vs the forecast at the time we had to use it. So our commercial colleagues send us a forecast 8 weeks out to allow us to buy components and raw materials. So when we get the final order 2 weeks out, we compare it to the forecast at 8 weeks and play the results back to them. For info we report both accuracy and bias, as we find different markets tend to have different bias. (Some over forecast, while other under forecast)

We are increasingly trying to be less and less reliant on forecasts, using them more to adjust inventory policy than to do the ordering (certainly for runners and repeaters). For our 'strangers' our purchase requirements tend to be well below MOQ for our suppliers anyway, so we work a re-order point process.