Assortment planning is a hot topic, especially amongst retailers, wholesalers, and the software developers that offer solutions to these industries. Yet despite lots of conversation, we hear very little discussion about the various clustering methodologies that lie at the heart of most assortment planning approaches. Parker Avery would like to help remedy that situation by examining the various clustering methodologies that we’ve encountered through working with a variety of retailers, with the aim of providing some insight into which technique or combination of techniques makes the most sense for your business model.
Before we dig into clustering, we need to briefly discuss assortment planning. Assortment planning is a term that has been in widespread use throughout the industry, yet does not have a clear, consistent definition. The meaning can vary depending on the perspective of the user and the situation. The term has been used variously to mean quantifying SKU-level sales and purchases, developing targeted assortments, assortment / space optimization, and more.
In looking at industry and academic literature dealing with assortment planning ranging over 40 years, nearly every aspect of merchandise planning and space planning has been included. One early attempt at definitive work on the subject1 included the design of the product hierarchy and layout of the display space as part of the scope of assortment planning. Clearly, that definition is too broad for this current exercise.
For purposes of this conversation, we will define assortment planning as “the practice of developing different assortments for targeted groups of customers.” There still may be other functions of the assortment plan. It may, for instance, be used to quantify purchases for each item or help determine the amounts of inventory to be distributed to each store and held back for direct sales. Yet, for this discussion the primary purpose of assortment planning is the development of tailored assortments.
Following this definition, clustering is the mechanism that is used to develop those targeted groups of customers. The ideal state of assortment planning would allow the targeting of a collection of products to each individual customer, based on his or her particular preferences. We may eventually be able to deliver on this ideal state through digital channels, but for the foreseeable future it will not be attainable in the bricks-and-mortar or catalog channels. This is because a multitude of customers and customer types patronize any individual store location, making individual targeting impossible. Clustering seeks to overcome this challenge by grouping together sales outlets (stores, website, catalog recipients, etc.) that demonstrate similarities in customer shopping behavior.
We can now turn our attention more fully to the topic of clustering. The term clustering refers to “the process of grouping sales outlets together based on similarities or patterns in their underlying customers’ behavior.” These similarities are most often gleaned from data related to historic or forecasted sales, or information that is descriptive of the customers or the store. Examples of the latter include demographic or climatic information.