Matches in SemOpenAlex for { <https://semopenalex.org/work/W768111943> ?p ?o ?g. }
- W768111943 abstract "In the latest decades, marketing has known a remarkable evolution. Whereas previously, product managers had the largest responsibilities when devising marketing actions, nowadays, the focus on products has shifted largely to a focus on customers. One of the main triggers of this evolution was undoubtedly the belief that it is several times less demanding – i.e. expensive – to sell an additional product to an existing customer than to sell the same product to a new customer. Following this reasoning, companies have increasingly focused on nurturing the customer-company relationships, and the term Customer Relationship Management (CRM) soon became one of the central topics in popular and academic marketing literature. Concurrently (and thanks to this evolution), a large diversity of companies have succeeded in building enormous transactional customer databases. These databases often register all interactions between a company and its customers, such as purchasing and complaint behavior, information requests, etc, and hence they allow companies to create a unique and coherent (“360°”) view of every individual customer. In a predictive modeling framework, this abundance of information is used predict future customer behavior at an individual customer level. Hence, the historical data can be used eg to predict whether a customer will purchase additional products from a given company, whether he/she will react to targeted promotions, whether he/she will be interested in certain products, whether he/she will spend more or less in the future, but also e.g. whether the customer will be able to refund a credit application. In sum, it can be stated that the applications of predicting individual customer behavior are mainly directed towards the domains of targeted marketing and consumer credit scoring. It should be clear that this type of analyses can be used to allow companies to service their customers in an cost effective, targeted manner. Using predictive techniques, companies succeed in producing relevant offes to all individual customers, and this is often what the – increasingly critical – modern customer expects. This dissertation features a large variety of applications in modeling individual customer behavior. More specifically, different studies focus on predicting customer loyalty, future spending, (partial) customer defection, targeting, and consumer credit scoring. Empirical results were given for companies active in retail, mailorder, telecommunications and banking. In this work, we distinguish five phases in the modeling process: (1) project definition, (2) creation of the table of analysis, (3) model building, (4) model validation and (5) implementation and model usage. Throughout six different studies, contributions were made to each of the five phases of the modeling process. The first two studies focus on the comparison of the predictive performance of different techniques stemming from either data mining or statistical backgrounds, being linear and logistic regression, linear and quadratic discriminant analysis, decision trees (C4.5 & C4.5 rules), Bayesians networks, ARD neural networks en random forests. In this selection, we have specifically investigated those techniques combine good predictive performance with a high user interpretability. The applications of these two studies focus respectively on predicting the future spending pattern and customer loyalty. In this section, we conclude among other things that traditional statistical techniques often prove to be reliable in predicting individual customer behavior, given that ample consideration is given to correct model validation, and the reduction of overfitting and multicollinearity by applying variable selection techniques. In a third study, the authors evaluate the usefulness of rewarding customers according to customer loyalty. The degree of an individual customer’s store loyalty is often unknown to companies, as they often only know the spending level of their customers at their own stores, but only rarely possess spending information at competitive stores. Based on a limited sample of customers, the authors succeeded to predict customer loyalty for all customers of a retailer, whereby this information can be used eg as a reward criterion in a loyalty program. In this study, we prove that such a reward program would succeed better in targeting those customers who deliver high word-of-mouth, and who exhibit high purchase intentions, and who are highly price insensitive. A fourth study is aimed at a specific problem in customer credit scoring. When a new credit score is constructed, only the outcome of previously accepted orders is available to credit scoring analysts, whereby the model should – by definition – be constructed on a sample of customers that is not representative of the future through-the-door applicant population. Due to the unique properties of the available data set, we were able to assess the impact of this bias on credit scoring performance and profitability. In this study, we conclude that this impact is significant yet modest. In the two latter studies of this doctoral dissertation, we investigate the validity of our findings across different industries and applications of targeted marketing and consumer credit scoring. In a fifth study, we show that the use of a random split of the data in training and validation set can be responsible for creating a large instability of the results. While this effect is often claimed to be non-existant when using large data sets, we note the opposite. Additionally, we offer a methodology that has the potential to reduce te variability of the split with a factor 800. In the sixth and last study, we evaluate the use of different variable selection procedures, and we investigate the cost (in terms of predictive performance) of increasing the face validity of a predictive model. In this study, we conclude that the predictive ability does not decrease when we ensure that all parameter signs of the final model correspond to the univariate parameter signs, whereby the acceptability of the predictive model is increased towards managers, employees and customers. Curriculum Vitae Geert Verstraeten (°Bonheiden, 1977) received the degree of Licentiaat in de Toegepaste Economische Wetenschappen at the Katholieke Universiteit Leuven in 2000. In 2001, he graduated from the Master in Marketing Analysis and Planning at Ghent University. In September 2001, he joined the Marketing Department at Ghent University as an assistant and project coordinator of the Master of Marketing Analysis. During that period, he performed his doctoral research aimed at predicting individual customer behavior. His studies appeared in European Journal of Operational Research, Journal of the Operational Research Society and Expert Systems with Applications. A fourth study is currently in the second review round of the Journal of Marketing. Geert has presented his research at different international scientific conferences in Europe, the US, Canada, and Latin America. Additionally, together with dr. Wouter Buckinx, their research was recently awarded the prize of SAS 2005 Student Ambassador in Lisbon, Portugal. Currently, Geert Verstraeten and Wouter Buckinx have established their own consulting firm that specializes in predicting individual customer behavior, under the brand name Python Predictions. Their website (www.pythonpredictions.com) provides more information on the ambitions and competences of the company and both entrepreneurs. For more information, please contact info@pythonpredictions.com." @default.
- W768111943 created "2016-06-24" @default.
- W768111943 creator A5028224109 @default.
- W768111943 date "2005-01-01" @default.
- W768111943 modified "2023-09-24" @default.
- W768111943 title "Issues in Predictive Modeling of Individual Customer Behavior: Applications in Targeted Marketing and Consumer Credit Scoring" @default.
- W768111943 cites W119778216 @default.
- W768111943 cites W1515026043 @default.
- W768111943 cites W1522114398 @default.
- W768111943 cites W1523293200 @default.
- W768111943 cites W1524313545 @default.
- W768111943 cites W1536719366 @default.
- W768111943 cites W1554983662 @default.
- W768111943 cites W1560979929 @default.
- W768111943 cites W1564211670 @default.
- W768111943 cites W1570448133 @default.
- W768111943 cites W1571998446 @default.
- W768111943 cites W1585547680 @default.
- W768111943 cites W1585743408 @default.
- W768111943 cites W1588059493 @default.
- W768111943 cites W1588952863 @default.
- W768111943 cites W1594420557 @default.
- W768111943 cites W1594945480 @default.
- W768111943 cites W1602329118 @default.
- W768111943 cites W1625504505 @default.
- W768111943 cites W1629179136 @default.
- W768111943 cites W1631913055 @default.
- W768111943 cites W1817561967 @default.
- W768111943 cites W1912123407 @default.
- W768111943 cites W1964378344 @default.
- W768111943 cites W1965680834 @default.
- W768111943 cites W1969087177 @default.
- W768111943 cites W1969423031 @default.
- W768111943 cites W1969430465 @default.
- W768111943 cites W1973849164 @default.
- W768111943 cites W1973948212 @default.
- W768111943 cites W1974478886 @default.
- W768111943 cites W1975694814 @default.
- W768111943 cites W197587573 @default.
- W768111943 cites W1976144991 @default.
- W768111943 cites W1979295864 @default.
- W768111943 cites W1979907230 @default.
- W768111943 cites W1982120517 @default.
- W768111943 cites W1983710941 @default.
- W768111943 cites W1986024564 @default.
- W768111943 cites W1988065538 @default.
- W768111943 cites W1997335181 @default.
- W768111943 cites W1997819647 @default.
- W768111943 cites W2003271439 @default.
- W768111943 cites W2005213340 @default.
- W768111943 cites W2005510983 @default.
- W768111943 cites W2005755239 @default.
- W768111943 cites W2008876898 @default.
- W768111943 cites W2012444665 @default.
- W768111943 cites W2015406385 @default.
- W768111943 cites W2019800008 @default.
- W768111943 cites W2026710079 @default.
- W768111943 cites W2029794145 @default.
- W768111943 cites W2033931717 @default.
- W768111943 cites W2034009564 @default.
- W768111943 cites W2036207307 @default.
- W768111943 cites W2039860729 @default.
- W768111943 cites W2040933675 @default.
- W768111943 cites W2041844120 @default.
- W768111943 cites W2043324736 @default.
- W768111943 cites W2043453086 @default.
- W768111943 cites W2048489440 @default.
- W768111943 cites W2048997388 @default.
- W768111943 cites W2052462336 @default.
- W768111943 cites W2058755489 @default.
- W768111943 cites W2060855982 @default.
- W768111943 cites W2063978378 @default.
- W768111943 cites W2063988370 @default.
- W768111943 cites W2064065622 @default.
- W768111943 cites W2066060244 @default.
- W768111943 cites W2067755834 @default.
- W768111943 cites W2069885022 @default.
- W768111943 cites W2070094080 @default.
- W768111943 cites W2071552263 @default.
- W768111943 cites W2076313627 @default.
- W768111943 cites W2076903334 @default.
- W768111943 cites W2077575809 @default.
- W768111943 cites W2078915716 @default.
- W768111943 cites W2079775628 @default.
- W768111943 cites W2088388649 @default.
- W768111943 cites W2093015399 @default.
- W768111943 cites W2094439750 @default.
- W768111943 cites W2097321609 @default.
- W768111943 cites W2098738227 @default.
- W768111943 cites W2099043863 @default.
- W768111943 cites W2100471135 @default.
- W768111943 cites W2101782549 @default.
- W768111943 cites W2102097552 @default.
- W768111943 cites W2102150307 @default.
- W768111943 cites W2102467277 @default.
- W768111943 cites W2112722199 @default.
- W768111943 cites W2113214579 @default.
- W768111943 cites W2115296209 @default.
- W768111943 cites W2116753260 @default.
- W768111943 cites W2117352691 @default.