Our client is a leading provider of aftermarket supply-chain management services in the Aerospace and Defense space. The primary business of the client was selling a wide range of aviation parts and tools. They did not have a systematic way to identify cross-sell opportunities resulting in lost revenue.
The client wanted to build a product recommendation platform that would recommend parts that a customer likely needs.
- Tens of thousands of aircraft spare parts being purchased in various quantities
- Some customers had rich historical data and others had limited data
- Could not use one single algorithm, and had to use a hybrid approach
- Developed a model to understand customer behavior leveraging their past purchase data
- Leveraged user-based collaborative filtering and association rule mining to develop recommendation models
- Estimated various models and evaluated precision at various recommendation levels. Precision is the % of recommendations that the customer would actually be interested in
- The recommendations platform is estimated to increase revenue by $ 6.4 MM annually for the specific segment we were working with