Our Client
A leading apparel retailer in Mainland China 
 
 
General Overview
Although traditional systems such as Enterprise Resource Planning (ERP) Software are able to provide retailers with basic insights such as “Top 10 Products”, “Top 10 Fast moving products etc, these insights are more descriptive and do not provide retailers with the next best actions they should take to improve sales.
 
By leveraging data science models, retailers will be able to split their outlets/stores into segments so that product assortments, size allocations, and promotional offers can be localized as needed. When embedded into different processes, retailers are able to do better forecasting, assortment planning, size-optimization, promotions planning, markdown optimization and replenishment scheduling. 
 
 
Objective
Our client was targeting 10X growth in terms of market share by leveraging data science into their current retail process. By deriving an optimal mix for their products based on consumer preferences would allow them to optimize the inventory cycle and improve overall sales. 
 
 
Solution Overview
Our client had a total of 300 stores and by leveraging machine learning algorithms, we were able to group these stores based on similarity of customer preferences within the store. These clusters are selected based on a silhouette score which is derived by the similarity a store is to its own cluster (cohesion) in comparison with other clusters (separation).
 
 
Solution Details 
The following table has been simplified for illustrative purposes: 
Levi Outcome
Once the feature mapping of the products was done, we run the model to identify the optimal number of clusters. The output of the statistical model is defined by the silhouette score that measures the separation and cohesion of the clusters.