Software

Jul 30, 2003. Salford Systems won all four first place awards in the Duke University/NCR Teradata center for CRM churn modeling tournament. The open competition asked participants to develop the best possible statistical/data mining model to predict churn for a major wireless telecommunications provider. The data, including summary data for 100,000 accounts, were extracted from the company's CRM data warehouse and made available to more than 30 international participants. The predictive models were evaluated for accuracy and top-decile lift on both contemporaneous and future data for separate unseen accounts.

Salford Systems won the competition by submitting the most accurate models overall and also by providing models with the greatest lift in the top decile of the "scorecard". Salford's tool of choice was TreeNet(tm), its most recently released data mining technology. Treenet builds on Salford's flagship product, the CART(R) decision tree, but develops "ensemble" models constructed from many (possibly thousands) of individual trees.

For more information, see www.salford-systems.com