A Practical Data Science Application: Developing Prediction Models for Product Inventory Reduction and Ongoing Monitoring to Create Efficiency

This study was developed for a client maneuvering through a global pandemic environment, resulting in lower demand and revenue. The client wanted to use this period as an opportunity to reduce inventory on several product lines. Research has established that there are applications of data science methods and algorithms in inventory management. This study seeks to determine if applied analytical models can help in a way that can reduce decision-making time and effort necessary for future product reduction efforts. This study applies multiple methods to various client-provided data to determine if those methods are a plausible fit for the production environment.

To test the hypothesis that data analysis methods can help identify $1M in product discontinuation opportunities, finished good product data was extracted and analyzed using various methods such as an artificial neural network (ANN) with cross-validation, principal components analysis (PCA), k-means clustering, and generalized linear modeling. A sample from 762 products was drawn and used to fit an ANN model for one product group. A test of the fitted model showed a low misclassification rate of 11.2% with the cross-validation analysis of the 3-node ANN fit.

These results show that an ANN analytical method across all product groups using product managers’ input shows strength in predicting future product reduction opportunities. Thus, the client should integrate the ANN modeling method into production data systems across the organization. Furthermore, features extracted from PCA methods can aid in additional endeavors.