Estimation of Joint Distribution from Marginal Totals with Applications to Database Marketing

In many customer surveys and database marketing applications such as segmentation, profiling and predicting consumers’ choices, the joint distributions of covariates of interests are required. However, for variety of reasons, such as protection of clients’ confidentiality, the data are often available only in marginal frequency distribution format. This makes it difficult to use the covariates in a meaningful way where joint distributions are required for analysis decision making. Availability of such joint distributions would enable the client, Bell Canada, to define customer segments, improve classification and predictive models and enhance the Company’s understanding of customers’ behaviour. This research will enhance Bell’s flexibility to define customer segments according to marketing needs and generate marketing profiles for segments of interest thereby improving their understanding of customers’ behaviour and allowing for development of customer-focused strategies.

Faculty Supervisor:

Dr. Fassil Nebebe

Student:

Debaraj Sen & Jordie Croteau

Partner:

Bell University Laboratories

Discipline:

Business

Sector:

Information and communications technologies

University:

Concordia University