Creation of a customer experience measure – an application of MARS
When a customer problem has just been solved, a service provider is keen to know how the customer is feeling about the company. A common way is to ask the customer to give an explicit feedback on a 0-10 scale which is used to calculate an NPS score. However this feedback is received only for a very small percentages of the cases. What to think of the silent customers? The VSSB analytics team has created a scoring solution, which uses internal data gathered about the resolution process and provides a measure for the customer experience. The Multiple Adaptive Regression Splines methodology performed very well despite of the poor data quality and delivered an acceptable solution quickly. Regression trees and outlier capping also contributed to the creation of a final successful measure, which is computable for any collection of the cases, resembling somewhat the NPS. But the biggest merit goes to the MARS, which provides a fast way to fit models in the presence of nonlinear variable relationships.
Data science team leader, Vodafone SSC Budapest
Tamás Molnár is working as the leader of the Group BI data science team located in Vodafone Shared Service Center Budapest. As such, he is interested not only in building models, but also the practicality of the results and finding possible use cases in the enormous Vodafone organization. By qualification he is applied mathematician and economist. Earlier Tamás held positions at Hungarian Telekom, Raiffeisen Bank and IFUA Horváth & Partners as data scientist and BI consultant.