Advanced Planning and Optimization Part 3: Store Clustering

Advanced Planning addresses the major functions supporting retail merchandising activities. An overview of these Advanced Planning functions was given in our first article  and they are shown here with today's topic highlighted.


Definition and Purpose
Store Clustering is the grouping of stores based on common store and demographic characteristics. There are two types of store clusters: performance and non-performance based. Performance based store clusters are grouped according to how they perform. For example, store locations with similar sales performance would be placed in the same store group. Non-performance based clusters consider store characteristics such as climate, store size and/or store type etc. Non-performance based clusters also consider customer demographics such as ethnicity, income level, age group, fashion preference etc.

Store clusters are usually developed at a Department and/or Class level. Store Clustering is very important as retail chains begin to consider customer-centric merchandising and tailor their assortments based on localized customer need. Assortment quantities and merchandise mix will be defined based on the characteristics of the various store clusters.

Effective store clustering often considers a combination of at least two types of clusters For example, it is very common to consider both store size and sales volume together as a method of clustering. As shown in the chart below,  this approach allows the planner to consider various combinations that will impact the assortments and products provided. A small store with high sales volume will be merchandised quite differently than a small store with low sales volume. The result of good clustering is an improved ability to provide a customer-centric merchandise environment.

Store Clustering Process and Methods
An overview of the pre-season Store Clustering process is shown below.

The key functions are described below.
- Store Cluster Review and Analysis - for both performance and non-performance based clusters, it is necessary to review the existing clusters to define any problems, issues and changes. For example, the analysis can consider parameters such as sales by volume group, climate zone, store size or other cluster grouping. Based on the analysis and changes, new clusters will be defined.

- Performance Based Cluster Definition - the parameters and algorithms that will be utilized to calculate the clusters are defined for each product area (e.g. department and/or class). The cluster criteria will be established such as the volume group breakpoints. History and/or store plan values will be selected. 

- Non-Performance Based Cluster Definition - based on the analysis and the latest store demographics and characteristics, non-performance based clusters will be defined. Cluster groups such as store size, climate, ethnicity and/or various customer demographics will be selected. For each cluster group, the parameters associated with the group will be defined (e.g. small, medium, large and extra large store size). The criteria for each parameter also need to be considered (e.g. the size range that defines small, medium, large and extra large stores).

- Store Plans and Forecasts - to develop the final performance based clusters, the latest store plans and/or store forecasts are needed. Store plans will be reviewed and necessary changes identified. The latest store forecasts, store trends as well as merchandise plan updates will be applied and new store plans will be developed.

- Final Store Cluster Development - for performance based clusters, the various algorithms will be run and the clusters calculated (e.g. sales volume). The clusters are then reviewed. As necessary, the parameters and/or algorithms may be changed and the clusters revised. For non-performance based clusters, the stores will be placed in the appropriate cluster group (e.g. store size/climate) based on the parameters and criteria established. The performance and non-performance based clusters will then be combined and the final store clusters are reviewed and modified as necessary.

Timing and Organization Considerations
Store Clustering is performed by department and/or class on a seasonal basis or more often. For assortment planning, the clusters may be defined for each buying season (e.g. back-to-school, holiday etc.). For allocation, the clusters may be updated each time a new allocation is performed (e.g. monthly).

Store Clusters are developed as part of the planning and allocation activity. In some cases, the allocators and/or store analysts will develop the clusters. Store planners may also have responsibility for cluster development. Allocators, store analysts and/or planners may perform the on-going maintenance and updating. The planners, analysts and allocators work as a team in the cluster development process.

Automated Support and Interfaces
Successful store clustering requires new systems to support the process and methods defined. The Store Clustering system considers: 

- Maintaining history and plans
- Maintaining algorithms and parameters
- Creating and updating performance based clusters 
- Developing non-performance based clusters 
- Providing review and revision capabilities

Store Clustering systems are supported by Business Analysis, Merchandise Planning and Store Planning systems as well as Forecasting/Trending. Future automation will begin to consider Advanced Clustering approaches in which optimization and data mining will be utilized to define the most effective clusters.

Summary and Benefits
Store Clustering provides the basis to plan a diverse multi-location environment in an efficient and timely manner. Store Clustering provides grouping stores based on both performance (e.g. sales volume) and non-performance (e.g. store size) parameters. The store clusters support assortment planning and allocation as well as maintaining a customer-centric merchandise approach. See below to see the interaction of Store Clustering with the other Advanced Planning processes.
This ad will auto-close in 10 seconds