Greasing the Wheels:  A Produce Recommender System in the Manufacturing Sector

In the manufacturing sector, resource groups like Customer Service, Sales, Marketing, and Pricing face significant challenges collaborating to deliver customized experiences to customers and drive upsell opportunities due to the time, resource, and skillset constraints required to analyze customers’ data.  This study develops an automated Recommender System to identify products customers may need through the use of Similarity Scoring on the customer and product attributes as well as Association Rules using eight months worth of customer transactional data in the manufacturing sector.  The model optimizes product predictions using a bespoke scoring model and finds that the optimal mix of hyperparameters involves casting a wide net for customer similarity, but, for products, tightening similarity thresholds using Market Basket Analysis.  The results generate an acceptable mix of both accurate predictions and product candidates (False Positive errors) to use for upselling and generating conversations with customers.  Business implications for one real-world manufacturer implementing this model total over $10 million in incremental revenue.