Maximizing Online Revisiting and Purchasing: A Clickstream-Based Approach to Enhancing Customer Lifetime Value

2023 | Journal of Management Information Systems | Citations: 0

Authors: Jabr, Wael; Ghoshal, Abhijeet; Cheng, Yichen; Pavlou, Paul

Abstract: Online retailers are increasingly focused on maintaining a long-term relationshi ... Expand

Abstract: Online retailers are increasingly focused on maintaining a long-term relationship with customers, encouraging repeat visits rather than single-time purchases to increase customer lifetime value. To help retailers maximize the probabilities of customers' revisiting and purchasing, we develop a two-stage model to better characterize and predict these two fundamental customer activities. In the first stage, we characterize the propensity of a customer revisiting the retailer's website. In the second stage, we develop a stochastic model that predicts revisits while also incorporating individual customer heterogeneity in exerted search effort during repeated visits. This heterogeneity is based on individual customer preferences in the choice of consideration sets, product information, pricing, and the search environment. Using customer level clickstream data, we show that our approach is not only better at predicting repeat customer visits, compared to existing methods, but also explainable and managerially interpretable. Most importantly, using computationally efficient simulation-based prescriptive analytics, we leverage our modeling approach to propose practical intervention strategies that maximize the joint likelihoods of customers revisiting and purchasing at the individual customer level. Collapse

Topics: website price management mobile system accounting logistics management

Methods: simulation logistic regression proportional hazards model cluster analysis machine learning

How to increase sustainable engagement in the workplace through green IS: the role of instructional and motivational design features

2023 | European Conference On Information Systems | Citations: 0

Authors: Krath, Jeanine; Morschheuser, Benedikt; Korflesch, Harald von; Hamari, Juho

Abstract: ... Expand

Abstract: Collapse

Topics: green IS mobile application systems design user behavior Python

Methods: linear regression analysis field study survival function statistical hypothesis test theory development

How Lending Experience and Borrower Credit Influence Rational Herding Behavior in Peer-to-Peer Microloan Platform Markets

2023 | Journal of Management Information Systems | Citations: 0

Authors: Lowry, Paul Benjamin; Xiao, Junji; Yuan, Jia

Abstract: This paper analyzes the herding behavior that characterizes lenders' lending dec ... Expand

Abstract: This paper analyzes the herding behavior that characterizes lenders' lending decisions on a microloan platform and explains how rational herding behavior can resolve the information-asymmetry problem, which is a well-known reason for the failure of online microloan platforms. Using a set of panel data on individual lending decisions acquired from Paipaidai.com (PPDai), an online microloan platform, we examine the influence of the lending decisions of prominent, experienced lenders on novice lenders to identify rational herding behavior. Our empirical analysis demonstrates that rational herding behavior can in fact efficiently reduce lender loss from borrower defaults caused by limited information. Although it is typically assumed that herding behavior is irrational, we find that it can be rational in this context and can thus shed light on why PPDai has succeeded while most other microloan platforms have failed. Accordingly, we make three key contributions: 1) we use heterogeneous herding effects to empirically determine whether lenders' herding behavior on PPDai is rational based on observational learning; 2) we investigate the moderating effect of borrower credit and novice-lender experience on herding, and we leverage this heterogeneity in lender experience to better explain loan results; and 3) because PPDai publicly provides potential lenders with a transparent credit score—in contrast to platforms like Prosper.com, which leverage hidden proprietary credit information from Experian—we further analyze the credit composition of prominent lenders to better understand the crucial determinants of rational herding. In fact, our follow-up survival simulations indicate that without rational herding, the total number of successful PPDai loans would have decreased by around 46 percent during the study period—a finding that further underlines the crucial influence of rational herding and the unique contextual factors of PPDai that have fostered it. Collapse

Topics: peer-to-peer model decision making peer-to-peer lending crowdsourcing data breach

Methods: simulation survival function proportional hazards model longitudinal research hazard function

Theories: lemon market theory loss aversion

Depicting Risk Profile over Time: A Novel Multiperiod Loan Default Prediction Approach

2023 | Management Information Systems Quarterly | Citations: 0

Authors: Wang, Zhao; Jiang, Cuiqing; Zhao, Huimin

Abstract: With the rapid development of fintech, the need for dynamic credit risk evaluat ... Expand

Abstract: With the rapid development of fintech, the need for dynamic credit risk evaluation is becoming increasingly important. While previous studies on credit scoring have mostly focused on single-period loan default prediction, we call for a new avenue-multiperiod default prediction (MPDP)-to depict risk profiles over time. To address the challenges raised by MPDP, such as monotonic default probability prediction and complex relationship accommodation, we propose a novel approach, hybrid and collective scoring (HACS). We design a hybrid modeling strategy to predict whether and when a borrower will default separately through a default discrimination model and a default time estimation model, respectively, and synthesize them through a probabilistic framework. To accommodate various possible patterns of default time and measure the distribution of default probability over successive time intervals, we propose a joint default modeling method to train the default time estimation model. Empirical evaluations at the model (time-to-default prediction performance and discrimination performance) and mechanism (identifiability and discriminability) levels, as well as impact analyses at the application (granting performance and profitability performance) level, show that HACS outperforms the benchmarked survival analysis and multilabel learning methods on all fronts. It can more accurately predict time-to-default and provide financial institutions and investors better decisionsupport in granting loans and selecting loan portfolios. Collapse

Topics: online lending risk management decision making analytical information system

Methods: survival analysis case study machine learning design science design artifact

Theories: portfolio theory

Context-Aware Marketing Attribution Based on Survival Analysis

2023 | International Conference on Business Informatics | Citations: 0

Authors: Züllig, Kilian; Napirata, Stefan; Zimmermann, Steffen

Abstract: Companies increasingly invest in digital marketing channels to promote their pr ... Expand

Abstract: Companies increasingly invest in digital marketing channels to promote their products and services. While the expenditures for each marketing channel are known, the contribution of marketing channels to a successful conversion, and therefore the value they generate, is unknown, but highly relevant for strategic decision-making. In this paper, we develop a novel, context-aware additive hazard marketing attribution (CAHMA) model based on survival analysis to address this problem. In addition to channel-specific, time-decaying effects of marketing on the users' conversion rate, we control for the effects of contextual features, such as the device or country from which users interact with marketing channels. Based on a prototypical implementation, we demonstrate the model's applicability and evaluate it on real-world data from the industry. We find that CAHMA outperforms other models in terms of accuracy while offering unique interpretability of the results and hence, providing deep insights for practitioners into the effects of marketing. Collapse

Topics: marketing management iOS Android mobile system MapReduce

Methods: machine learning survival analysis hazard function design artifact Markov chain Monte Carlo

Theories: probability theory

In-Training Multicalibrated Survival Analysis For Healthcare via Constrained Optimization

2022 | International Conference on Information Systems | Citations: 0

Authors: Suttaket, Thiti; Kok, Stanley

Abstract: Survival analysis is an important problem in healthcare because it models the r ... Expand

Abstract: Survival analysis is an important problem in healthcare because it models the relationship between an individual's covariates and the onset time of an event of interest (e.g., death). It is important for survival models to be well-calibrated (i.e., for their predicted probabilities to be close to ground-truth probabilities) because badly calibrated systems can result in erroneous clinical decisions. Existing survival models are typically calibrated at the population level only, and thus run the risk of being poorly calibrated for one or more minority subpopulations. We propose a model called GRADUATE that achieves multicalibration by ensuring that all subpopulations are well-calibrated too. GRADUATE frames multicalibration as a constrained optimization problem, and optimizes both calibration and discrimination in-training to achieve a good balance between them. Empirical comparisons against state-of-the-art baselines on real-world clinical datasets demonstrate GRADUATE's efficacy. In a detailed analysis, we elucidate the shortcomings of the baselines vis-à-vis GRADUATE's strengths. Collapse

Topics: online learning system

Methods: survival function nonparametric test logrank test experiment survival analysis

Can Bot Improve Equity? Machine-generated Content Mitigates Cold-Start Issue

2021 | International Conference on Information Systems | Citations: 0

Authors: HOU, JINGBO; Chen, Pei-yu

Abstract: The cold-start problem is salient in the current online two-sided markets. Theo ... Expand

Abstract: The cold-start problem is salient in the current online two-sided markets. Theoretically, machine-generated content can serve as the information signal, thus helps to address the cold-start issue. To empirically examine whether machine-generated content helps mitigate the cold-start problem, we study the impact of machine-generated content on dataset adoption in a leading online public dataset community (Kaggle). We found that machine-generated content helps to solve the cold-start problem in the online public dataset community in two ways. First, we show that machine-generated content shortens the time needed for datasets to get their first adoption and that machine-generated content increases the dataset adoption in the initial stage and increases the dataset diffusion rate. Second, our results show that the positive effect of machine-generated content is stronger for dataset sharers who lack reputation signals. Our research opens a discussion on the role of machine-generated content in mitigating the cold-start problem in online public dataset platforms and also other two-sided markets. Collapse

Topics: word of mouth advertising management data breach

Methods: instrumental variables estimation survival analysis survival function robustness check statistical hypothesis test

Theories: lemon market theory signaling theory

Reviewing Before Reading? An Empirical Investigation of Book-Consumption Patterns and Their Effects on Reviews and Sales

2021 | Information Systems Research | Citations: 0

Authors: Lee, Heeseung Andrew; Choi, Angela Aerry; Sun, Tianshu; Oh, Wonseok

Abstract: Over the past decades, research on online book reviews has inundated academic ci ... Expand

Abstract: Over the past decades, research on online book reviews has inundated academic circles with numerous theoretical reflections and empirical manifestations aimed at explaining the effects of such resources on business performance. Yet, these studies succumbed to the conventional pitfall of assuming that consumers write reviews only after they fully read the book that they purchased. A recent industry report revealed that although many individuals initiate book reading, only a few finish them. With these considerations in mind, we investigated how consumers’ book-consumption patterns affect their review behaviors and how reviews generated from incomplete consumption influence subsequent sales. We used expectation confirmation theory (ECT) as a theoretical foundation to elaborate on the review behaviors of consumers at various stages of their e-book consumption. On the basis of ECT, we argue that customers can submit reviews not necessarily after full consumption, but at any point during this trajectory, and even when consumption has yet to take place. Consumption patterns were traced and captured from records of reading activities on e-book devices and apps. Our results indicate that a considerable number of consumers provide positive reviews even before initiating reading or after progressing up to an extremely early section of a book. In addition, the relationship between review valence and completion rate can be characterized as a U-shaped pattern, given that reviews arising from negative disconfirmations occur more frequently than those emerging out of positive disconfirmations. The findings also uncover that customers occupying the extreme ends of the completion continuum provide less extreme review ratings. The effect of completion rate on review length is significantly positive. Our text-analysis results suggest that reviews based on sufficient consumption contain more useful insights than do those grounded in incomplete consumption. Moreover, review comments formed after incomplete consumption adversely affect subsequent sales. Finally, we discuss a number of our findings’ implications and provide actionable recommendations that can aid platforms in their efforts to refine their online-review systems and policies in pursuit of enhanced credibility in peer evaluation. Collapse

Topics: ebook digital good online review user-generated content digital platform

Methods: natural language processing robustness check instrumental variables estimation logistic regression hazard function

Theories: expectation confirmation theory

On Estimation of Equipment Failures in Electric Distribution Systems Using Bayesian Inference

2021 | HICSS | Citations: 0

Authors: Peerzada, Aaqib; Begovic, Miroslav M.; Rohouma, Wesam; Balog, Robert

Abstract: This paper presents a new statistical parametric model to predict the times-to-f ... Expand

Abstract: This paper presents a new statistical parametric model to predict the times-to-failure of broad classes of identical devices such as on-load tap changers, switched capacitors, breakers, etc. A two-parameter Weibull distribution with scale parameter given by the inverse power law is employed to model the survivor functions and hazard rates of on-load tap changers. The resulting three-parameter distribution, referred to as IPL-Weibull, is flexible enough to assume right, left, and even symmetrical modal distribution. In this work, we propose an inferential method based on Bayes’ rule to derive the point estimates of model parameters from the past right-censored failure data. Using the Monte Carlo integration technique, it is possible to obtain such parameter estimates with high accuracy. Collapse

Topics: database system

Methods: Bayesian analysis hazard function survival function experiment

From Lurkers to Workers: Predicting Voluntary Contribution and Community Welfare

2020 | Information Systems Research | Citations: 13

Authors: Kokkodis, Marios; Lappas, Theodoros; Ransbotham, Sam

Abstract: In an online community, users can interact with fellow community members by volu ... Expand

Abstract: In an online community, users can interact with fellow community members by voluntarily contributing to existing discussion threads or by starting new threads. In practice, however, the vast majority of a community’s users (≈90%) remain inactive (lurk), simply observing contributions made by intermittent (≈9%) and heavy (≈1%) contributors. Our research examines increases and decreases of types of user engagement in online communities using hidden Markov models. These models characterize latent states of user engagement from trace user activity or lack of activity. The resulting framework then differentiates lurkers who can later become workers (i.e., engaged in the community) from those who will not. Differentiating lurkers who can be engaged from those who cannot enables managers to anticipate and proactively direct their resources toward the users who are most likely to become or remain workers (i.e., heavy contributors), thereby promoting community welfare. Analysis of 533,714 posts from an online diabetes community shows that incorporating latent user engagement variables can significantly improve the accuracy of welfare prediction models and guide managerial interventions. Application of our framework to five additional communities of various contexts demonstrates its generalizability. Collapse

Topics: online community participatory design user behavior social network pseudonymity

Methods: machine learning XGBoost deep learning logistic regression longitudinal research