A support vector machine (SVM) approach to imbalanced datasets of customer responses: comparison with other customer response models

Customer response is a crucial aspect of service business. The ability to accurately predict which customer profiles are productive has proven invaluable in customer relationship management. An area that has received little attention in the literature on direct marketing is the class imbalance problem (the very low response rate). We propose a customer response predictive model approach combining recency, frequency, and monetary variables and support vector machine analysis. We have identified three sets of direct marketing data with a different degree of class imbalance (little, moderate, high) and used random undersampling method to reduce the degree of the imbalance problem. We report the empirical results in terms of gain values and prediction accuracy and the impact of random undersampling on customer response model performance. We also discuss these empirical results with the findings of previous studies and the implications for industry practice and future research.     

Customer base analysis with recurrent neural networks

One of the primary goals that researchers look to achieve through customer base analysis is to leverage historical records of individual customer transactions and related context factors to forecast future behavior, and to link these forecasts with actionable characteristics of individuals, managerially significant customer sub-groups, and entire cohorts. This paper presents a new approach that helps firms leverage the automatic feature extraction capabilities of a specific type of deep learning models when applied to customer transaction histories in non-contractual business settings (i.e., when the time at which a customer becomes inactive is unobserved by the firm). We show how the proposed deep learning model improves on established models both in terms of individual-level accuracy and overall cohort-level bias. It also helps managers in capturing seasonal trends and other forms of purchase dynamics that are important to detect in a timely manner for the purpose of proactive customer-base management. We demonstrate the model performance in eight empirical real-life settings which vary broadly in transaction frequency, purchase (ir)regularity, customer attrition, availability of contextual information, seasonal variance, and cohort size. We showcase the flexibility of the approach and how the model further benefits from taking into account static (e.g., socio-economic variables, demographics) and dynamic context factors (e.g., weather, holiday seasons, marketing appeals). We make an open-source reference implementation of the newly developed method available at https://github.com/valendin/rfm2lstm.     

Next-product-to-buy models for cross-selling applications

We present and evaluate next-product-to-buy (NPTB) models for improving the effectiveness of cross-selling. The NPTB model reduces the waste of poorly targeted cross-selling activities by predicting the product each customer would be most likely to buy next. We describe the model-building process and discuss theoretical and practical issues in developing a NPTB model. We then illustrate the effectiveness of the NPTB approach with a field test. The field test shows that the NPTB model increases profits compared to a heuristic approach, and that profits are incremental over and above sales that would have occurred through other channels. We then conduct an empirical test of methodological issues. We find that incorporating current product ownership as a predictor enhances predictive accuracy the most, followed by customer monetary value to the company, and demographics. We find that statistical method makes little difference in predictive accuracy, with neural nets having a slight edge. A simple random sample to create the calibration database increases predictive accuracy more than a stratified random sample, although the stratified sample may be preferred to avoid underpredicting unpopular products. We explore the potential for incorporating purchase incidence models in the NPTB approach, and find that this potentially enhances the effectiveness of the NPTB model. We close with recommendations for practitioners and for future academic research.     

Understanding the Impact of Online Reviews on Customer Choice: A Probability Discounting Approach

This study uses the concept of probability discounting to understand the impact of online customer reviews on consumer choice. Probability discounting describes how the subjective value of an outcome alters when its delivery shifts from certain to uncertain. An experimental study with 29 participants was conducted. Participants were run through an online shopping scenario where they had to choose whether to buy a product from a Web shop with customer reviews on reliability or from a Web shop without reviews but with a lower product price. A titration procedure over sales price for the Web shop without reviews was run over seven probability conditions. The mean switching points where participants chose where to buy the product were extracted from the experimental data, and probability discounting factors were calculated. The results supported the assumption that online reviews indicate the probability of a successful transaction online and function as a guide to choices. Implications for marketers as well as suggestions for future research are discussed.     

The impact of personalised incentives on the profitability of customer retention campaigns

Traditional approaches to managing customer churn have typically concentrated on those customers most likely to defect. While accurately predicting customer churn probability is important, this metric alone does not sufficiently empower managers to make optimal decisions. Hence, the current study focuses on the relationship between retention incentives and profit maximisation. Specifically, we improve existing churn management practices by: (1) allowing for customer heterogeneity in incentive redemption behaviour, (2) introducing the dependence of the probability of accepting an incentive on its monetary value, and (3) offering an improved model for developing retention campaigns. We support our conclusions with empirical data and simulations and make tangible managerial recommendations.     

2008 DMEF Customer Lifetime Value Modeling (Task 2)

We describe our approach for predicting individual donor's total gift amount over a two-year target period. We divide the donors into 8 segments; for each segment, we fit a logit model for predicting the probability of giving, and a log-linear model for predicting the amount of gifts conditional on a donor giving. We found that recency, frequency, and first gift amount are good predictors of the probability of giving, while time-weighted total gift amount in the past years is a good predictor for future gift amount.     

Customer Lifetime Value as the Basis of Customer Segmentation

SUMMARY

In this article, we examine current trends in customer life-time value and customer segmentation models and identify key issues for future research. CLV-based segmentation is a segmentation approach that groups customers into meaningful segments based upon customer lifetime value and (potentially) other factors. In the article, we discuss the extent to which CLV-based segmentation meets the criteria for effective segmentation. We also identify six areas for future research: (1) models and management of “micro-segments,” (2) using CLV-based segmentation to improve the efficiency of marketing programs, (3) the need for more dynamic CLV-based segmentation models, (4) applying CLV-based customer segmentation to new products and new customers, (5) challenges associated with implementing CLV-based segmentation, and (6) the need for new models that enable firms to segment customers by response to marketing activities and CLV at different points in the customer decision process.     

Customer Experience Creation: Determinants, Dynamics and Management Strategies

Retailers, such as Starbucks and Victoria's Secret, aim to provide customers a great experience across channels. In this paper we provide an overview of the existing literature on customer experience and expand on it to examine the creation of a customer experience from a holistic perspective. We propose a conceptual model, in which we discuss the determinants of customer experience. We explicitly take a dynamic view, in which we argue that prior customer experiences will influence future customer experiences. We discuss the importance of the social environment, self-service technologies and the store brand. Customer experience management is also approached from a strategic perspective by focusing on issues such as how and to what extent an experience-based business can create growth. In each of these areas, we identify and discuss important issues worthy of further research.     

Stochastic service life cycle analysis using customer reviews

This study proposes a stochastic service life cycle analysis to gauge where a service is in its life cycle and to give forecasts about its future prospects. We employ customer review data to measure customer-oriented service maturity and use a hidden Markov model to estimate the probability of a service being at a certain stage of its life cycle. Based on this, we also develop three indicators to represent the future prospects of a service’s life cycle progression. The main advantages of the proposed approach lie in its ability to model different shapes of life cycles without any supplementary information and to examine a wide range of services at acceptable levels of time and cost. We believe our method will assist firms in building stage-customised post-launch service strategies. A case study of mobile game services in the Apple App Store is presented.     

Counting your customers from an ��always a share�� perspective

Models based on the Pareto/NBD framework are among the most popular for customer base analysis. The Pareto/NBD framework assumes that purchasing follows a Poisson process until the customers defect. Therefore, models based on this framework may systematically underestimate the number of future transactions from customers whose probability of returning is greater than zero. In this paper, we propose a new model which assumes that customers do not defect, but instead switch freely between an active and an inactive state. We call this model the ??interrupted Poisson process??. According to the model, customers purchase through a Poisson process when they are active and they do not purchase when they are inactive. Bayesian simulation methods for parameter estimation are developed and implemented via a Markov chain Monte Cacrlo (MCMC) simulation. Several useful expressions for customer base analysis are derived. Through simulation experiments, we show that the rate of customers moving from an inactive to an active state is an important factor determining the fit and predictive ability of the Pareto/NBD model and our model. An empirical analysis, using two real-life datasets, demonstrates the superior performance of the proposed model.     

Constructing a relationship-based brand equity model

The purpose of this study is to develop and test a model in which several aspects of the service encounter including service staff, servicescape, customer similarity, and customer interaction are taken into account simultaneously as antecedents of relationship quality and generation of brand equity. Testing the hypotheses involved two service settings, banks and department stores. The findings demonstrate that serviced staff and customer interaction have significant direct effects on brand equity. Surprisingly, four variables of service encounter have significant indirect effects through relationship quality on brand equity. Based on these findings, the implications for managers and future research are identified.     

The impact of online customer satisfaction on the yahoo auction in Taiwan

In this article, the authors develop a conceptual model that links website quality, trust, merchandising, customer service, and online satisfaction for yahoo auction in Taiwan. The research objective is to provide initial evidence for the determinants of e-satisfaction (online satisfaction). We examine the role of web quality, trust, merchandising, and customer service in consumer online satisfaction assessments. This conceptual model is empirically tested from 350 consumers across a broader group of online shoppers on yahoo auction by means of internet surveys and structural equation analysis. The results show that the determinants of online trust are website quality, merchandising, and customer service. Website quality, trust, merchandising, and customer service have positive effects on online satisfaction for auction sites. The research findings were reported by discussing the implications of the findings and directions for future research.     

A Partially Hidden Markov Model of Customer Dynamics for CLV Measurement

Customer lifetime value (CLV) measurement is challenging as it requires forecasting customers' future purchases. Existing stochastic CLV models for this purpose generally make the following assumptions: 1) purchase behavior of customers can be described by purchase frequency and the average monetary value of transactions, 2) customers keep the same purchase behavior pattern over time, 3) purchase frequency and monetary value are independent, and 4) customers are active during a limited period of time after which they permanently defect. We develop a new stochastic model that relaxes these four assumptions. First, in addition to the number of transactions and its monetary values, we also model purchase incidence decisions (i.e. whether or not to purchase). Second, our partially hidden Markov truncated–NBD-GG (PHM/TNBD-GG) model allows dynamic purchase patterns, dependence between purchase frequency and monetary value, and customers to become active after a few periods of temporary inactivity. Validation of our model on two datasets demonstrates that if assumptions 1 to 4 of existing stochastic models are violated our model produces more accurate forecasts of future customer behavior.     

服务行业顾客重购意向的影响因素及其相互关系研究

本文探讨了转换壁垒、顾客感知价值、顾客满意对顾客重购意向的影响作用及其相互关系。以理发行业为研究对象,通过发放问卷收集数据,并运用结构方程分析软件进行了实证性检验,结果发现:顾客满意和顾客感知价值都对顾客重购意向具有直接显著影响;转换壁垒的不同维度对顾客重购意向的影响不同,社会利益对顾客重购意向产生积极的影响作用,转换成本不能增加顾客价值和顾客满意,但会对顾客产生锁定作用。对企业来说,管理者可以增加顾客满意和顾客感知价值来增加顾客重购意向,也可以通过提高顾客对转换壁垒的感知,从而对顾客起到锁定作用。     

Transformation of Metrics and Analytics in Retailing: The Way Forward

The retail industry is undergoing tremendous changes that are driven by technology, changing consumer tastes, economic pressures, competition, stakeholder relationships, environmental concerns, and governmental regulations. Our article explores analytics as a capability that helps retailers excel in this dynamic environment. We identify the reasons behind the trends in the retail industry and provide guidance for retail managers on how to improve customer relationship management using appropriate metrics and effective analytics. Our guidance to retail managers emphasizes the importance of brand recognition, explores tactics for enhancing customer experience, recommends establishing superior customer engagement, forging social connections among consumers, and rendering service and support to customers, and highlights a data-oriented approach to retailing. We conclude with suggestions for future research in this domain.     

CRM in Data-Rich Multichannel Retailing Environments: A Review and Future Research Directions

Many retailers have collected large amounts of customer data using, for example, loyalty programs. We provide an overview of the extant literature on customer relationship management (CRM), with a specific focus on retailing. We discuss how retailers can gather customer data and how they can analyze these data to gain useful customer insights. We provide an overview of the methods predicting customer responses and behavior over time. We also discuss the existing knowledge on the application of marketing actions in a CRM context, while providing an in-depth discussion on CRM and firm value. We outline future research directions based on the literature review and retail practice insights.     

The Influence of Cognitive Dissonance on Retail Product Returns

Given the significant costs and customer service ramifications associated with the return of retail merchandise it is important to understand the underlying reasons for product returns. One such underlying reason is cognitive dissonance. Customers who experience cognitive dissonance may seek to undo the effects of a regretted choice by returning the product in question. This research examines the influence of two forms of cognitive dissonance (emotional dissonance and product dissonance) on the frequency of product returns. Three antecedents (consideration of liberal return policies, customer opportunism, and switching barriers) are examined in terms of their influence on cognitive dissonance and product returns. In addition, the moderating role of gender and store brand is reported. The research is based on a survey of Wal‐Mart and Target customers who engaged in product returns. Structural equation modeling is used to verify and test these relationships. Emotional dissonance and product dissonance were found to be positively related to product returns frequency. It was found that consideration of liberal return policies reduces both emotional and product dissonance, while customer opportunism and switching barriers increase both dimensions of cognitive dissonance. Both gender and store brand were found to be significant moderators of the relationships between cognitive dissonance and two antecedents (consideration of liberal return policies and customer opportunism). In addition, gender and store brand moderated the linkage between product dissonance and emotional dissonance, and the linkage between emotional dissonance and return frequency.     

The 11 Questions Every Donor Asks and the Answers All Donors Crave: How You Can Inspire Someone to Give Generously,by H. McKinnon

Organizational learning and customer orientation have been a focus of research for a number of years in both marketing and management literature. Customer learning orientation is conceptualized as three important components: management customer orientation, customer feedback, and employee learning orientation. By drawing from both marketing and organizational research theories, the authors propose a model of customer learning orientation in a public sector organizational setting. Customer learning orientation is hypothesized to have a significant effect on employee attitudes of role ambiguity and self-efficacy, which in turn affects job outcomes of job satisfaction and organizational citizenship behaviors. Using a sample of 438 employees of a public sector organization, the authors test the model through a structural equation modeling technique. The results provide general support for the model. Implications for managers of public sector organizations and future research are discussed.     

Will the frog change into a prince? Predicting future customer profitability

More and more companies have customer databases that enable them to analyze customer profitability over time. These companies often seek to determine the most important customers as indicated by their current or historical profitability and focus attention on them. Focusing on profitable customers can result in more efficient use of marketing resources, but this approach neglects the fact that customers can evolve over time. Some customers begin as low-profit customers but eventually develop into high-profit customers. Others may start out as high-profit customers but become unprofitable over time. Previous efforts to predict future profitability have been relatively unsuccessful, with relatively simple, naïve models often performing just as well as or better than more sophisticated ones. Our paper presents a new approach to predicting customer profitability in future periods that performs significantly better than naïve models. We estimate the models on data from a high-tech company in a business-to-business context and validate the models' predictive ability on a holdout sample.We show that a model based on simulation of customer futures provides large improvements over naïve extrapolation of average profits. By using the simulation model to select customers, ROI from marketing efforts is projected to increase by 58%.