Online Program

Client Attrition Analytics at IBM
View Presentation View Presentation *Subrata Chatterjee, IBM 
Dion Cummings, IBM 
Spyros Kontogiorgis, IBM 
Helene Miller, IBM 

Keywords: customer attriton, Cox regression, Propotional hazards model, B2B customer erosion

Identifying customers that are likely to attrite along with the factors behind their attrition are critical elements in sustaining organic growth, increasing profitability as well as market share. Also, identifying the customers who are most likely to switch to another vendor allows optimization of sales and marketing intervention resources that are crucial for a large multinational company such as IBM. The Advanced Client Analytics group at IBM has built several models based on the discrete version of Cox Proportional hazards model to accurately identify customers that are most likely to attrite or decrease their spending with IBM brands. The models are able to flag accounts that are likely to attrite in the next two years even though there are no signs of attrition in the revenue stream or opportunity pipeline. IBM sellers are using the customer-level erosion scores and the factors behind the attrition to propose specific solutions that are most likely to resonate with the business model of their clients. Erosion scores are also being combined with the opportunity pipeline information to gain additional insights around the likelihood of winning specific deals for IBM brands.