Syntha Lab RU

Store clusters: productivity and mix profile

Groups stores by the productivity index (sales per m² to the network average) and by the similarity of the category profile to the network profile. The result is clusters A, B, D and the “special profile” flag for the size curve, display and replenishment.

Assortment and sizes For whom: Chain managers and category managers who split stores into clusters for the curve, display and replenishment.

Calculate on your data →

At a glance

What question does it answer?
Which stores are productive per m², which are weak and which have a category profile different from the network.
What data is needed?
Area and sales of stores for one period.
How to read the result?
The productivity index sets the cluster, the profile similarity marks special stores. Read new stores separately. Example on the default values: Network sales per m² (mature stores): RUB 299,125; Stores in cluster A: 2; Stores in cluster B: 1; Stores in cluster D: 1. The figures are illustrative: enter your own to see your result.

Why and how to reason

Stores with one level of sales per m² and a similar category profile can be run by one curve and one display. The productivity index is sales per m² to the network average, the profile similarity is the cosine of category shares. In the example there are four stores: indices 1.23, 0.91, 0.61, 1.29, clusters A, B, D, A, store 3 has a special profile.

When to use it

Apply once a season before the size curve, display and replenishment. It is a grouping, not the distribution of goods across stores (cluster).

How to read the result

The productivity index sets the cluster, the “special profile” flag says the common curve does not suit. Read new stores separately.

Common mistakes

Clustering is by total sales, new stores are mixed with mature ones, too many clusters are made.

What data is needed?

  • Area and sales of stores for one period.
  • Category shares of the store.
  • The age of the store.

Inputs

FieldUnitValue in the exampleMeaning
Productivity index of core A: from×1.2Stores with an index not below this value go to cluster A. The thresholds are set by you for the network, there are no norms. The example value is illustrative.
Productivity index of cluster D: below×0.8Stores with an index below this value go to cluster D, between the thresholds it is cluster B. The example value is illustrative.
Minimum profile similarity to the network×0.98Cosine similarity of category shares; below the threshold the store is flagged “special profile” and needs its own curve and display. The example value is illustrative.
Minimum age of a mature store, monthsmo12Younger stores are excluded from the network average and profile: their sales are still growing. Empty in the table means mature. The example value is illustrative.

How it is calculated

StepFormula
Network sales per sqmthe sum of sales of mature stores ÷ the sum of their areas
Productivity indexPI = (store sales ÷ sqm) ÷ network sales per sqm
Network profilecategory shares weighted by store sales
Profile similaritycos = sum(a·b) ÷ (root sum a² × root sum b²), a and b: category shares of the store and the network
ClusterA at PI not below 1.2, D at PI below 0.8, otherwise B; a flag at cos below 0.98

Worked numbers

Default values from the app; in the interactive version you enter your own.

Network sales per sqm (mature stores)
299,125 ₽
Stores in cluster A
2
Stores in cluster B
1
Stores in cluster D
1
Stores with a special profile
1
Sales share of cluster A, %
62.6 %
Sales share of cluster D, %
16.8 %
Productivity index of stores
Bar chart: Store 1 1.23, Store 2 0.91, Store 3 0.61, Store 4 1.2900.511.5Store 1Store 1Productivity index: 1.23Core A threshold: 1.2Cluster D threshold: 0.8Store 2Store 2Productivity index: 0.91Core A threshold: 1.2Cluster D threshold: 0.8Store 3Store 3Productivity index: 0.61Core A threshold: 1.2Cluster D threshold: 0.8Store 4Store 4Productivity index: 1.29Core A threshold: 1.2Cluster D threshold: 0.8
  • Productivity index
  • Core A threshold
  • Cluster D threshold
Data as a table
MeasureProductivity indexCore A thresholdCluster D threshold
Store 11.231.20.8
Store 20.911.20.8
Store 30.611.20.8
Store 41.291.20.8

Store clusters

StoreSales per sqmProductivity indexProfile similarityClusterNote
Store 1368,077 ₽1.230.985A—
Store 2272,778 ₽0.910.989B—
Store 3183,182 ₽0.610.861Dspecial profile
Store 4387,143 ₽1.290.981A—

Stores Store 3 have a special profile: the common size curve and display of the network do not suit them.

There are fewer than six stores: the clusters are conditional. For a network of up to 50 points do not make more than five or six clusters.

Region, climate and format (a mall or a street) often set the cluster stronger than the area; compare the result with them.

What to do with the result: build the size curve (sizedemand), display and replenishment by cluster and separately for a special profile; recount the clusters once a season.

A typical mistake: clustering by total sales instead of per sqm, mixing new stores with mature ones and making too many clusters.

Nuances and limits

  • Clustering by total sales rather than per m².
  • Mixing new stores with mature ones.
  • Making too many clusters.

Terms used

Calculate on your data →