Sharad Borle

Customer Lifetime Value Measurement

The measurement of customer lifetime value is important because it is used as a metric in evaluating decisions in the context of customer relationship management. For a firm, it is important to form some expectations as to the lifetime value of each customer at the time a customer starts doing business with the firm, and at each purchase by the customer. In this paper, we use a hierarchical Bayes approach to estimate the lifetime value of each customer at each purchase occasion by jointly modeling the purchase timing, purchase amount, and risk of defection from the firm for each customer. The data come from a membership-based direct marketing company where the times of each customer joining the membership and terminating it are known once these events happen. In addition, there is an uncertain relationship between customer lifetime and purchase behavior. Therefore, longer customer lifetime does not necessarily imply higher customer lifetime value. We compare the performance of our model with other models on a separate validation data set. The models compared are the extended NBD--Pareto model, the recency, frequency, and monetary value model, two models nested in our proposed model, and a heuristic model that takes the average customer lifetime, the average interpurchase time, and the average dollar purchase amount observed in our estimation sample and uses them to predict the present value of future customer revenues at each purchase occasion in our hold-out sample. The results show that our model performs better than all the other models compared both at predicting customer lifetime value and in targeting valuable customers. The results also show that longer interpurchase times are associated with larger purchase amounts and a greater risk of leaving the firm. Both male and female customers seem to have similar interpurchase time intervals and risk of leaving; however, female customers spend less compared with male customers.

Conjugate Analysis of the Conway-Maxwell-Poisson Distribution

This article explores a Bayesian analysis of a generalization of the Poisson distribution. By choice of a second parameter , both under-dispersed and over-dispersed data can be modeled. The Conway-Maxwell-Poisson distribu- tion forms an exponential family of distributions, so it has sucien t statistics of xed dimension as the sample size varies, and a conjugate family of prior distribu- tions. The article displays and proves a necessary and sucien t condition on the hyperparameters of the conjugate family for the prior to be proper, and it discusses methods of sampling from the conjugate distribution. An elicitation program to nd the hyperparameters from the predictive distribution is also discussed.

A Model of the Joint Distribution of Purchase Quantity and Timing

Prediction of purchase timing and quantity decisions of a household is an important element for success of any retailer. This is especially so for an online retailer, as the traditional brick-and-mortar retailer would be more concerned with total sales. A number of statistical models have been developed in the marketing literature to aid traditional retailers in predicting sales and analyzing the impact of various marketing activities on sales. However, there are two important differences between traditional retail outlets and the increasingly important online retail/delivery companies, differences that prevent these firms from using models developed for the traditional retailers: (1) the profits of the online retailer/delivery company depend on purchase frequency and on purchase quantity, whereas the profits of traditional retailers are simply tied to total sales, and (2) customers in the tails of the frequency distribution are more important to the delivery company than to the retail outlet. Both of these differences are due to the fact that the delivery companies incur a delivery cost for each sale, whereas customers themselves travel to retail outlets when buying from traditional retailers. These differences in costs translate directly into needs that a model must address. For a model intended to be useful to online retailers, the dependent variable should be a bivariate distribution of frequency and quantity, and the frequency distribution must accurately represent consumers in the tails. In this article we develop such a model and apply it to predicting the consumers joint decision of when to shop and how much to spend at the store. Our approach is to model the marginal distribution of purchase timing and the distribution of purchase quantity co nditional on purchase timing. We propose a hierarchical Bayes model that disentangles the weekly and daily components of the purchase timing. The daily component has a dependence on the weekly component, thereby accounting for strong observed periodicity in the data. For the purchase times, we use the Conway-Maxwell-Poisson distribution, which we find useful to fit data in the tail regions (extremely frequent and infrequent purchasers).

Open source software success: Measures and analysis

Despite a growing body of research on OSS production, much remains to be learned. One important issue concerns the measures of OSS project success and its determinants. In this paper, we empirically study the determinants of OSS success as measured by the number of subscribers and developers working on an OSS project. Furthermore, we demonstrate that our model forecasts these success measures more accurately as compared to a naive model. We find that OSS projects that develop software to work on Windows/UNIX/Linux operating systems, and developed using C or its derivative languages experience larger increase in subscribers and attract more developers than projects that do not have these characteristics. OSS projects with semi-restrictive licenses have fewer subscribers and attract fewer developers. Interestingly, OSS projects that accept financial donations and are targeted at IS/IT professionals have more subscribers than others, although these characteristics do not affect the developer base. The number of subscribers and developers increases with the age of the OSS project. Finally, the impact of developers on subscribers and subscribers on developers is positive and significant.

A generalized framework for estimating customer lifetime value when customer lifetimes are not observed

Measuring customer lifetime value (CLV) in contexts where customer defections are not observed, i.e. noncontractual contexts, has been very challenging for firms. This paper proposes a flexible Markov Chain Monte Carlo (MCMC) based data augmentation framework for forecasting lifetimes and estimating customer lifetime value (CLV) in such contexts. The framework can be used to estimate many different types of CLV models—both existing and new. Models proposed so far for estimating CLV in noncontractual contexts have built-in stringent assumptions with respect to the underlying customer lifetime and purchase behavior. For example, two existing state-of-the-art models for lifetime value estimation in a noncontractual context are the Pareto/NBD and the BG/NBD models. Both of these models are based on fixed underlying assumptions about drivers of CLV that cannot be changed even in situations where the firm believes that these assumptions are violated. The proposed simulation framework—not being a model but an estimation framework—allows the user to use any of the commonly available statistical distributions for the drivers of CLV, and thus the multitude of models that can be estimated using the proposed framework (the Pareto/NBD and the BG/NBD models included) is limited only by the availability of statistical distributions. In addition, the proposed framework allows users to incorporate covariates and correlations across all the drivers of CLV in estimating lifetime values of customers.

Estimating the Contextual Risk of Data Breach: An Empirical Approach

Abstract Data breach incidents are on the rise, and have resulted in severe financial and legal implications for the affected organizations. We apply the opportunity theory of crime, the institutional anomie theory, and institutional theory to identify factors that could increase or decrease the contextual risk of data breach. We investigate the risk of data breach in the context of an organization’s physical location, its primary industry, and the type of data breach that it may have suffered in the past. Given the location of an organization, the study finds support for application of the opportunity theory of crime and the institutional anomie theory in estimating the risk of data breach incidents within a state. In the context of the primary industry in which an organization operates, we find support for the institutional theory and the opportunity theory of crime in estimating risk of data breach incidents within an industry. Interestingly though, support for the opportunity theory of crime is partial. We find that investment in information technology (IT) security corresponds to a higher risk of data breach incidents within both a state and an industry, a result contrary to the one predicted by the opportunity theory of crime. A possible explanation for the contradiction is that investments in IT security are not being spent on the right kind of data security controls, a fact supported by evidence from the industry. The work has theoretical and practical implications. Theories from criminology are used to identify the risk factors of data breach incidents and the magnitude of their impact on the risk of data breach. Insights from the study can help IT security practitioners to assess the risk environment of their firm (in terms of data breaches) based on the firm’s location, its industry sector, and the kind of breaches that the firm may typically be prone to.

Common Value vs. Private Value Categories in Online Auctions: A Distinction Without a Difference?

There is a growing body of empirical research that attempts to distinguish private value from common value auctions. Strategic behavior for a seller/bidder in these two paradigms should differ, so the assumption of, or the identification of, the type of auction (private or common) is important for understanding the auction dynamics and strategies in game-theoretic models. However, it is difficult to recognize which of the two paradigms applies to a particular good. In this article, we briefly review some of the empirical work distinguishing common/private values using observed bidding behavior, both structural nonparametric and parametric models of auctions. We then examine the use of a priori beliefs to classify product categories. Specifically, we survey auction experts on their subjective judgment in classifying a list of consumer product categories on a private/common value continuum. We also survey consumers, asking them their subjective assessment. Interestingly, not only are extant models unable empirically to distinguish between private value and common value auctions, but also the experts have strongly divergent opinions on classifying product categories on this continuum. These findings raise doubts about the appropriateness of using a priori beliefs to classify a product as a common value versus a private value, and furthermore the findings question the feasibility of game-theoretic models to empirically distinguish private value from common value auctions. Perhaps decision theory would be a more useful paradigm for modeling auction decisions in practice.

Does Membership in a Nominal Online Group Affect Long-Term Customer Behaviors? Results from a Naturalistic Field Experiment on Kiva.org

In this exploratory investigation, we examine changes associated with joining a nominal online group on longer-term behavior of lenders on the leading peer-to-peer microcredit site Kiva.org. Using a random sample of 5,000 Kiva lenders and exploiting a naturalistic field experimental manipulation (the introduction of the “lending teams” feature on Kiva), we analyzed the relationship of group membership with lending behaviors over a two year period. Our results reveal that lending team membership is positively associated with both the number of loans granted and the amount of money loaned whereas non-membership is associated with drop-offs in both behavioral measures. These results suggest that providing customers with the opportunity to join nominal online groups based on their existing affiliations (such as their alma mater, favorite sports team, nationality, etc.) may offer marketers with a relatively inexpensive, low-maintenance, and effective method of strengthening relationships with customers.

The Impact of Facebook Fan Page Participation on Customer Behavior: An Empirical Investigation

We report the results of a longitudinal field study, conducted in cooperation with two restaurants, to examine the degree to which participating on a Facebook fan page affects customer behaviors. We assessed customers’ baseline levels of visit frequency and spending. The restaurants then set up Facebook fan pages, maintaining them with regular updates of interesting marketing content. The same customer pools were invited to participate on the restaurants’ respective Facebook fan pages, and five months later, were re-surveyed to reassess their behaviors. Relatively small percentages of invited customers (approximately 2%) became the restaurants’ Facebook fans. Contrary to conventional wisdom, becoming a Facebook fan led to decreased spending per visit, whereas the visit frequency remained unchanged. We also found selective mere measurement effects of survey participation. Customers participating in the first survey visited the restaurant more frequently. The study’s findings raise doubts about the effectiveness of Facebook fan pages in providing direct benefits to firms from the fans in terms of their spending and visit frequency. However, there is some evidence of indirect benefits (e.g. new customer acquisition) that firms can gain from their Facebook fans.