Understanding Customers' Behaviour through Web Data Mining Models

Introduction

The main challenge for companies is to identify the best models able to forecast marketing performance, especially in today’s competitive landscape. In fact, a fundamental part of managing organizations is planning future market trends. Nowadays, the long-run success of corporations is closely related to how well management is able to anticipate future market tendencies and draw up competitive marketing and business strategies. According to Sato (2000) data mining analysis differs from the statistical data analysis. In fact, statisticians study the population parameters drawing observation sample by estimation, testing and predictions with the main risk to forecast wrong future market trends. For this reason, the development of predictive data mining models is becoming one of the main priorities for managers in every industry. Data mining analysis has become an astonishing approach for marketers given that the meaningful knowledge is often hidden in huge customer database and most traditional statistical methods could fail to uncover such knowledge. Bueren et al. (2004) define customer knowledge as a strategic intangible asset useful to obtain a competitive advantage in the markets. An efficient utilization of the latter determines the competitive development of corporations. This is particularly true in marketing field due to the dramatic proliferation of e-customer data available online. Indeed, companies have acknowledged that their marketing strategies should focus on identifying those customers who are likely to churn in order to minimize the churn risk probability and maximize the probability of customer conversion in the short, medium and long run (Hadden, Tiwari, Roi & Ruta 2005). Churn management is defined as a set of techniques that enable corporations to keep their profitable customers in order to increase customer loyalty (Lejeune, 2001).

Prediction of behaviour, customer value, customer satisfaction and customer loyalty are example of some of the information that can be extracted from the data that should already be stored within a customer’s database. In fact, one of the main tools able to help marketers in the mentioned approach is data mining due to the large amount of data available in this database (Khak & Glolamain, 2010). Finally, the key goal for businesses, regardless their size, is to discover strategic knowledge within large database in order to minimize the probability of customer’s risk of churn.

The purpose of this research is to demonstrate the strategic importance of web data mining models in estimating the probability of customer’s risk churn in competitive landscape. Special attention is paid on General Linear Models (GLMs) such as Logistic Regression Models based on ‘Enter Methods’. The Criteria based on the Loss Functions such as the Confusion Matrix and the Receiver Operating Characteristic (ROC) curve have been used to evaluate the goodness of fit of the predictive models developed in this work.

This chapter is further structured as follows. Section 1 provides a theoretical framework of the main difference between statistical data analysis and data mining analysis due to the big lack in the literature. Section 2 describes the research methodology implemented in this research as well as the data collection process. Section 4 shows the main empirical findings with particular attention of the criteria based on the loss functions. Section 5 highlights conclusions. Finally, Section 6 shows the main future research directions.