بهینه سازی مدیریت ریزش با استفاده از متغیرهای بازاریابی قابل کنترل و مدیریت هزینه های مرتبط
|کد مقاله||سال انتشار||تعداد صفحات مقاله انگلیسی||ترجمه فارسی|
|5823||2013||10 صفحه PDF||سفارش دهید|
نسخه انگلیسی مقاله همین الان قابل دانلود است.
هزینه ترجمه مقاله بر اساس تعداد کلمات مقاله انگلیسی محاسبه می شود.
این مقاله تقریباً شامل 8650 کلمه می باشد.
هزینه ترجمه مقاله توسط مترجمان با تجربه، طبق جدول زیر محاسبه می شود:
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Expert Systems with Applications, Volume 40, Issue 6, May 2013, Pages 2198–2207
In this paper, we propose a churn management model based on a partial least square (PLS) optimization method that explicitly considers the management costs of controllable marketing variables for a successful churn management program. A PLS prediction model is first calibrated to estimate the churn probabilities of customers. Then this PLS prediction model is transformed into a control model after relative management costs of controllable marketing variables are estimated through a triangulation method. Finally, a PLS optimization model with marketing objectives and constraints are specified and solved via a sequential quadratic programming method. In our experiments, we observe that while the training and test data sets are dramatically different in terms of churner distributions (50% vs. 1.8%), four controllable variables in three marketing strategies significantly changed through optimization process while other variables only marginally changed. We also observe that the most significant variable in a PLS prediction model does not necessarily change most significantly in our PLS optimization model due to the highest management cost associated, implying differences between a prediction and an optimization model. Finally, two marketing models designed for targeting the subsets of customers based on churn probability or management costs are presented and discussed.
The propensity of customers to terminate their relationships with service providers has forced many companies in competitive markets to shift their strategic focus from customer acquisition to customer retention (Chen and Hitt, 2002 and Venkatesan and Kumar, 2004). This is mainly because companies can increase the average net present value of a customer by up to 95% by boosting the customer retention rates by 5%. In particular, with exceptionally high annual churn rates (20–40%), the mobile telecommunications service providers eager to launch successful churn management programs to maximize their revenues (Kim et al., 2004, Eshghi et al., 2007 and Glady et al., 2009). For this purpose, many data mining and statistical models have been presented to accurately identify prospects or possible churners in the automotive, insurance, and telecommunication industry. Such models include PLS regressions (Lee, Kim, Lee, Cho, & Im, 2011), decision trees (Kim, 2006 and Xiea et al., 2009), ANNs (Buckinx and Poel, 2005 and Mozer et al., 2000), support vector machines (Coussement & Van den Poel, 2008), genetic algorithms (Au et al., 2003 and Kim et al., 2005), or dynamic programming models (Gönül & Shi, 1998). A recent discussion about advantages and disadvantages of various models for churn management can be found in (Hadden et al., 2007 and Neslin et al., 2006). While such prediction models are very important for successful churn management, most prediction models are limited in the sense that they do not consider implementations costs associated with churn management programs (Bult and Wansbeek, 1995 and Kumar and Shah, 2004). Note that, according to these predictive models, marketing managers may identify likely churners based on the estimated churn probability, and chooses top x% of customers as target customers to offer retention marketing promotions (Lee et al., 2011). However, most retention marketing promotions bear different cost structures and hence should be administered with care. For example, one of the most common retention programs among telecommunication service providers is to provide customers who renew their contract periods a financial incentive that allows them to purchase a new mobile device at a deeply discounted price. Another popular retention program is to simply provide a better customer call center service in regards to billing and call quality peacefully through educated and experienced receptionists to resolve many questions and complaints and enhance customer satisfaction and loyalty ( Fornell and Wernerfelt, 1987, Gustafsson et al., 2005, Mittal and Kamakura, 2001 and Reinartz et al., 2005). Note that while two retention programs may or may not be equally effective, providing a new mobile device at a deeply discounted price may cost more than providing a better call center service. In this paper, we propose a churn management model based on partial least square (PLS) optimization that explicitly considers management costs of controllable marketing variables. The PLS method in this paper will be used not only as a prediction model to predict churners but also as a control model combined with optimization method to maximize the effects of churn management strategies at minimum cost. Ideally, limited resources for retention promotions should be allocated to most likely churners who generate most revenues while minimizing the management costs of such retention promotions. The detailed objectives of this research are: (1) to categorize and validate controllable and uncontrollable marketing variables; (2) to determine the management costs of each controllable marketing variable by applying a triangulation analogy method; and (3) to develop and solve a PLS optimization model that minimizes the total cost of implementing three retention marketing strategies while satisfying the objective of retention marketing strategies. The remainder of this paper is organized as follows. Section 2 provides a brief review of PLS model for prediction and control purposes and a triangulation method for management cost estimation. In Section 3, the overall research framework is introduced, and data sets are explained. In the following Section 4, a PLS-based optimization model is presented in a mathematical form after controllable marketing variables are identified and their management costs are assigned. Section 5 first presents experimental results from a PLS optimization model designed for entire customers. Then two marketing models designed for targeting the subsets of customers based on churn probability or management costs are presented and discussed. Finally, Section 6 provides the conclusion of the paper and suggests several direction of further research.
نتیجه گیری انگلیسی
In this paper, we present a churn optimization model based on the PLS prediction and control model. At first, a PLS regression model with chosen variables that meet a certain threshold is calibrated to estimate the churn probability of all the service users, and it is turned into an optimization model after dividing variables into controllable and uncontrollable variables, and associating management costs and other constraints with controllable variables. Then, the optimization problem is solved by using a SQP algorithm. The advantages of the proposed model are numerous. In terms of methodology, it significantly enhances the scalability and interpretability of a prediction model by including only predictive variables in its final model. In our example, it selects only 46 variables (out of 123 original variables) with high VIP scores and then constructs six latent variables of these 46 variables for final prediction. In addition, it allows marketing managers to explicitly develop a marketing campaign as a mathematical optimization model and efficiently solve it using a SQP algorithm. In particular, the proposed model and solutions are found to be robust to the dramatic changes in churner distributions of training and test data sets. From the perspective of marketing managers, the proposed model is very customizable and generalizable by allowing managers to incorporate different marketing strategies and associated marketing variables in each marketing strategy. In particular, marketing managers can not only chose appropriate and preferred strategies (e.g., CMS, RMS, and DMS in this study) and controllable variables to meet their unique needs, but also apply their subjective weights and management costs to control variables. By incorporating a triangulation method to estimate relative management costs of each controllable variable, the proposed method not only allows different marketing managers to use different cost schemes but also the outcomes of the proposed model can be still applicable to different scenarios as long as the relative ranking of management costs for controllable variables is remained the same. Finally, the managerial and financial outcomes of the proposed model can be discussed at two levels. First of all, the proposed model can reduce the churn probability of all the customers in the database while minimizing the cost of marketing campaign. The reduced churn probability of all the customers in turn implies that customers are less likely to terminate their current service contract and hence the service provider enhances the profitability over the life time of customers. Another advantage of the proposed model is that it allows marketing managers to accrue additional financial profits by limiting the scope of marketing campaigns to a subset of customers based on estimated churn probability or management cost. One of limitations of the proposed model is that management costs associated with marketing variables do not directly bear the accounting or financial values and, hence, it is not straightforward to estimate the financial values of marketing campaigns. Therefore, a follow-up study may estimate the impact of input variables when they are associated with absolute financial values on the outcomes of the proposed model. In particular, it will be interesting to estimate the financial values from reduced churn probability of all the customers in the optimization process for a fixed structure of management costs of controllable marketing variables while changing the objective goal (e.g., how much should we reduce the churn probability of customer). At the same time, additional financial profits from customer selected based on churn probability or management cost can also be estimated. By doing so, marketing managers can not only determine an optimal objective goal in the optimization process but also maximize the financial profits for target marketing.