For a large set of SKUs and in two successive selling seasons, we have compared the accuracy of three quantitative forecasting methods based on advance (preview) demand information. The methods all use a top-down approach, and ?rst forecast the aggregate total demand for a group of SKUs by scaling up the aggregate preview demands. They differ in the subsequent division of that aggregate forecast over the individual SKUs; proportional to preview demand (Method 1), equal (Method 2), or top-?op.
A ?rst result was that all top-down methods performed best when the top level was de?ned at the more detailed product group level rather than the broader assortment group level or even for all assortment groups together. This is a useful ?nding for the case company, as they currently use the assortment group level. Amongst the considered top-down methods, the so-called top ?op method turned out to have the most robust performance. This method is based on the assumptions that, within a group of related SKUs, the top (?op) SKUs always account for a certain fraction of total demand for that group. To the best of our knowledge, this method has not been tested or described in the literature before.
Our results certainly suggest that it would be interesting to test the method for other real-life cases as well. An especially attractive feature of the top-?op method, as appeared from the overall results and was illustrated for speci?c product groups, is that it avoids over-forecasting for those SKUs with the largest preview demands. This is done by 'spreading' the total preview demand per class equally amongst the SKUs in a class, rather than dividing based on the individual preview demands per SKU as the more 'straightforward' Method 1 does.
Over-forecasting leads to too large orders and obsolescence problems, where left over stock needs to be discarded or sold below the cost price at the end of the season. For one assortment group with 89 SKUs, we extended the comparative study to include expert judgment methods. It turned out that these methods outperformed all other methods. However, Tukey tests showed the performance differences not to be signi?cant at the 5% level. So, we should be cautious in announcing these expert judgment methods the 'winners'.