Unexpected Customer Relationship Leads to Doubling Market Share Through Predictive Analytics and Data Mining

• Activity Number: 159 - SPEED: Sports and Business
• Type: Contributed
• Date/Time: Monday, July 31, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Steven Reagan*
• Companies: L&L Products, Inc.
• Keywords: Price Prediction ; Artificial Neural Network ; MATLAB ; Sales/Price Forecasting ; Customer relationship building ; Out-of-sample prediction

Abstract

This presentation will walk the audience through a learning path of unexpected relationships, research, and implementation of solutions to data mining challenges experienced while predicting corporate sales performance. Specifically, the speed and accuracy with which a component price can be delivered were found to have such value to global customers that our market share could be doubled. This presentation will highlight an instance of price modeling using [very] early, design stage, information, previously thought of little value, to deliver pricing quotes within 10% of final offering - with no RFQ, engineering feasibility studies, or Sales quotation process involvement! Artificial neural networks are developed using up-front design information and augmented with bootstrap aggregation techniques for model robustness. These models were ultimately deployed onto desktop environments at customer installations. Additionally, a new covariance stable method of out-of-sample test data generation will be shown that was critical in the selection of reliable models.

The modeling tools involved were MATLAB, JMP (SAS), and R.