An apparel brand redesigning its product hierarchy should write down what division, class, style, color, size and season each mean, and have finance, planning and merchandising sign that document, before its new planning system is configured.
A study in the journal Entropy, published April 24, 2019 by researchers at INESC Technology and Science, the University of Porto and the Polytechnic Institute of Porto with no external funding, states that to ensure aligned decision-making across the hierarchy, it is essential that forecasts at the most disaggregated level add up to forecasts at the aggregate levels above. The study forecast along the hierarchical structure of products adopted by the retailer, from store through area, division, family, category and sub-category down to SKU, using weekly sales from one of the largest stores of a Portuguese supermarket chain between January 2012 and April 2015. The fit to apparel is loose. The authors kept only products with at least one sale every week, a filter that a style sold for one season would fail.
Forecasting: Principles and Practice, a textbook by two Monash University professors in its third edition since 2021, sets the requirement for forecasts to add up in a manner that is consistent with the aggregation structure; the two most common top-down methods specify disaggregation proportions based on the historical proportions of the data; and attributes that are crossed rather than nested, such as a bicycle manufacturer's frame size or price range, do not naturally disaggregate in a unique hierarchical manner.
In apparel, division, class and style nest, so each style belongs to one class, and a style moved to a new class should take its sales history with it, since the common top-down proportions are based on history. Color and size sit under style, and the document should decide when a new colorway of an existing style counts as a color rather than a new style. Season crosses that chain much as frame size does; the document has to state whether a style carried into a second season keeps its first season or takes the new one. Finance should sign the levels at which it sets the merchandise budget, because the forecasts below those levels will add up to them.
In weeks one to four the head of merchandising drafts one definition for each level and attribute, with an example style. Through weeks five to eight the planning director restates past seasons' sales under those definitions. From week nine the FP&A lead ties the restated class totals to the sales finance reported for the same periods, and in week thirteen the three sign and configuration begins.
The work ends at the signed document and the restated history. Which forecasting method runs, statistical or machine learning, stays with the implementation team, because the textbook shows any set of base forecasts can be made to add up through the hierarchy. A later change to any definition returns to the same three signatures.
A brand that sells through retailers is also asked to send its color data outside the company. GS1 US, the not-for-profit information standards organization that acquired the National Retail Federation's color and size code tables in 2020 and sells them through its store, tells vendors in a guideline released June 29, 2020 to put the selected color in the product master data feed to trading partners, and says retailers then extract this information for their merchandise database. The same guideline lets a vendor use one color code for an entirely different shade on another style, and GS1 US recommends that vendors not assign its color codes in their internal systems.
