Store Clustering
Store clustering groups selling locations that respond to the same assortment in the same way, so one plan can serve a group of doors instead of one plan per door.
What is store clustering?
Store clustering is the practice of grouping selling locations that respond to the same assortment in the same way, so that one plan can be built for the group rather than one plan for every door. A cluster is a planning unit rather than a description of a store: doors belong together when giving them the same option list, the same depth profile and the same size or shade distribution would produce the same outcome.
The cluster decides what a group of doors receives — which options, which sizes, shades or price bands, and when they arrive. Allocation then decides how much each door inside the cluster receives, sized on its own rate of sale and its own space. Clustering exists because both alternatives fail at scale: one assortment everywhere repeats the same mistake in every door, and one plan per door multiplies the planning work by the door count while the evidence under each plan gets thinner.
How store clustering works
Clustering evaluates doors across a small number of dimensions, and the useful ones are those that change a decision:
- Sales velocity: average units sold per option per week, which sets how much evidence a cluster can produce
- Price sensitivity: full-price versus markdown sell-through, which tells you where a door's price ladder actually sits
- Size distribution: the shape of the size curve a location sells, fitted only on the weeks it held a complete run
- Climate and seasonality: the week demand for a category historically opens, which moves flow dates before it moves content
- Trade-area composition: the customer profile behind the demand, and the attribute used when a door has no usable history of its own
The output is a set of cluster assignments, each receiving its own option list, depth profile and size or shade distribution. Allocation then sizes each door inside the cluster on its own rate of sale and its own space.
Why store clustering matters
Without clustering, brands default to one of two extremes: sending the same assortment everywhere, which repeats the same mistake in every door, or planning each door individually, which is operationally impossible at scale. Both produce inventory distortion — overstocked doors sitting alongside stocked-out ones.
Clustering improves sell-through rate by matching product to local demand, reduces markdown exposure by not over-allocating to low-velocity doors, and cuts the planning work from one plan per door to one plan per group. Clusters are reviewed on a fixed annual cadence, and doors move between them on structural changes — a remodel, a relocation, a change in selling space or channel — rather than on a single weak season.
In RetailNorthstar: the assortment is built at cluster level, locations are assigned to clusters, and allocation places inventory across the doors inside each cluster. Cluster definitions are configurable per category and per season, and clusters can be proposed from prior-season sell-through patterns. Choosing the clustering dimension and the cluster count stays with the planning team.
Where the full treatment is
The working method — choosing the clustering dimension, deciding how many clusters a fleet can carry, what to do when the history is missing or censored, how clusters feed allocation and size curves, the re-clustering cadence, and the six ways clusters break — is in the pillar guide:
- Store Clustering and Localization by Vertical — the cross-vertical treatment across apparel, footwear, accessories and bags, home and furniture, outdoor, sporting goods, health and beauty, toys and games, baby and juvenile, and jewelry and watches, including the dimension that carries the variation in each one.
- Store Clustering for Apparel Brands — the apparel deep dive: climate clusters for outerwear, cluster-level size curves, and DTC versus wholesale door clustering.
- Localized Assortment Planning — what happens after the cluster map exists, and how localization ties back to the financial plan.