Statistical Modeling And Analysis Based On Database Marketing From A Non-profit Organization

In the recent century, there is long evolution in guiding theory of western corporation marketing management, where several concepts are generated, such as production concept, products concept, promotion concept and marketing concept. As competition is more brutal, market segmentation is speeding up, information system cost is reduced, new concepts of marketing emerge and get developed, where database marketing is outstanding. In order to implement database marketing, corporations need to develop efficient modeling systems, thus scholars all around the world are adapting kinds of statistical and data mining techniques to build the typical model of database marketing, scoring model. However, there is still no agreement about which scoring model is most suitable, which is therefore what scholars and data analysts are most concerned about.This paper introduces the concept and importance of database marketing, and summarizes the most adapted statistical and data mining techniques in it. Furthermore, by introducing marketing data collecting, modeling and analysis process of a Non-Profit organization in United States, Na-tional Multiple Sclerosis Society (NMSS for short), this paper shows the significant role database marketing is playing inside corporation sales and customer relationship maintenance.This paper has deeply explored Scoring model, based on improving response rate for NMSS marketing events as the business objective. The details are as following:a) Response-revenue two stage Scoring model is proposedThis model is a two stage model combining logistic regression and linear regression, aiming to improve response rate and revenue per responder. Moreover, by employing statistical knowledge, model is built under a standard process, including data resampling, data transfer, variables reduc-tion, model selection and tuning, and model validation. Besides, variables reduction in this process is based on the correlation of variable, which is different from traditional method.b) Measure method of Scoring model based on Decile is developedApart from traditional measurement such as receiver operating characteristic curve, this paper adapts a novel method based on decile, which pays more attention to how customer is scored from the model. By ranking Scoring of every customer from the model and divide these records into ten deciles, a lift chart could be drawn, where more sloping the curve is, more efficient the model is when brought into production.c) Advice and recommendationAdvice and recommendation has been provided for NMSS and similar problems when carrying out database marketing, based on result of the built model.