Data Mining & Machine Learning to empower business strategy

October 7, 2004
Oliver Downs | Princeton University

This talk will focus on my recent work for the Acquisition and Retention team at MSN, in which we address the business problem of successfully predicting customer behavior far enough ahead that the predictions are actionable. The goals of this work were; to impact churn, by allowing intervention to retain valuable customers; to allow the monetization of customers leaving the MSN dialup service for Broadband, by targeting high upgrade propensity customers to partners, and focusing those living outside areas with a partner relationship on MSN’s BYOA (Premium) product.

Using the SQL Server 2005 data mining tools, we developed and deployed a set of 4 predictive models on MSN’s database of nearly 5 million active subscribers, around which MSN’s Acquisition & Retention team now builds business strategy.

During the course of this work, we uncovered a simple abstraction of the subscription relationship that can be applied to any contract-free subscription business, in particular the post-number-portability wireless industry.

Speaker Details

Olly Downs was one of MSR’s longest serving interns, and held the MSR Graduate Fellowship while in the PhD Program in Applied & Computational Mathematics at Princeton University. At MSR, together with Chris Burges, he conceived the basis of his PhD thesis work, Quantum Tunneling Optimization. In August 2002, Olly founded Analytical Insights, Inc. with the goal of applying his expertise in Machine Learning & Data Mining to applied problems, and promptly began working with MSN. This work lead to engagements with Novantas LLC, Apollo Data Technologies, American Express, AT&T Wireless, Regal Cinemas, The Seattle Times, the US Government and Oren Etzioni’s Hamlet Inc.Olly also holds BA, and MSci. Degrees in Experimental & Theoretical Physics, from the University of Cambridge, UK.