Demand forecast error and the capacity buffer

Why it matters

A demand forecast is the number produced at a given moment, with the information on hand, for a future need. Changing the forecast retroactively after actual demand is known makes the error smaller, but does not improve planning ability. For this reason, the date on which the forecast was frozen, and the delivery period and product scope it was frozen for, must be kept. It is possible to compare the last actual result of the same day with the forecast given that morning; comparing it with a forecast corrected at the end of the day answers a different question. The reliability of the error calculation depends on this match of time and scope before the subtraction. If different forecast horizons are combined without explanation, the result cannot be interpreted.

When we define the error sign as actual minus forecast, a positive value shows demand higher than expected and a negative value shows demand lower than expected. This sign is useful for tracking the direction of a capacity gap. But positive and negative errors can cancel each other when added. A mean signed error of zero does not prove that the daily plans were right. Absolute error preserves the size and shows the deviation in both directions together. The two measures are used side by side: one describes a persistent tendency to under- or over-forecast, the other the size of the fluctuation. Neither alone calculates delivery success; for that, the time sequence is needed.

A capacity buffer is resource room that can really be used above the planned load. This room is not just an empty hour in the schedule; the right equipment, skilled workers and material must be present together. Also, a buffer and stock are not the same thing. Acceptable product made earlier can meet demand, but in this lesson daily capacity is examined without opening stock. A day's unused time is not physically carried to the next day. If early production is not allowed, the slack on a low-demand day does not create ready product on the next busy day. For this reason the buffer decision is tested with a daily backlog calculation.

Using past errors does not mean assuming the future will be the same. The capacity that covers the largest error of a short series may be sufficient for that series; it cannot be said to provide a particular delivery probability. The product mix, the demand pattern and resource losses may change. The buffer decision is explained with these limits and retested with different demand series. The same average demand and the same absolute error can create different delays in different sequences. The aim is not to close off all uncertainty with a single percentage, but to make visible which deviation the decision covers and under what circumstances it will be revisited.

Formula