Search

Firms adopt randomized experiments to evaluate various interventions (e.g., website design, creative content, and pricing). However, most randomized experiments are designed to identify the impact of one specific intervention. The literature on randomized experiments lacks a holistic approach to optimize a sequence of interventions along the customer journey. Specifically, locally optimal interventions unveiled by randomized experiments might be globally suboptimal when considering their interdependence as well as the long-term rewards. Fortunately, the accumulation of a large number of historical experiments creates exogenous interventions at different stages along the customer journey and provides a new opportunity. This study integrates multiple experiments within the reinforcement learning (RL) framework to tackle the questions that cannot be answered by stand-alone randomized experiments. How can we learn optimal policy with a sequence of interventions along the customer journey based on an ensemble of historical experiments? Additionally, how can we learn from multiple historical experiments to guide future intervention trials? We propose a Bayesian recurrent Q-network model that leverages the exogenous interventions from multiple experiments to learn their effectiveness at different stages of the customer journey and optimize them for long-term rewards. Beyond optimization within the existing interventions, the Bayesian model also estimates the distribution of rewards, which can guide subject allocation in the design of future experiments to optimally balance exploration and exploitation. In summary, the proposed model creates a two-way complementarity between RL and randomized experiments, and thus, it provides a holistic approach to learning and optimizing interventions along the customer journey. This paper was accepted by Anindya Ghose, information systems. Funding: This work was supported by Adobe Faculty Research Award and the Marketing Science Institute Research Grant. Supplemental Material: The data files and online appendix are available at https://doi.org/10.1287/mnsc.2023.4914. [ABSTRACT FROM AUTHOR]

We propose an approach for uncovering characteristic response paths of a population from an individual-level multivariate time series data set. The approach is based on a model that accommodates a set of arbitrary distributions for endogenous variables and interstep intervals and variables. The model enables reliable estimation of individual-level parameters by uncovering and statistically pooling clusters of similar individuals. We show that using such a model one can distribute the response of an outcome variable to an impulse over all possible preceding activity sequences. When a few such sequences explain most of the response, they describe the population's characteristic response paths from the impulse to the outcome. We apply the proposed approach to a customer touchpoint data set from a large multichannel specialty retailer. This application uncovers six customer segments, each with unique characteristic paths to purchase. These paths provide insights into the behavior of customers and the optimal over-time communication strategy for different customer segments. Summary of Contribution: Uncovering users' paths through physical and virtual spaces has been of considerable interest in the computing and operations research domain. The existing research suggests a demand for visualizing the primary paths of agents through geographic, online, and activity spaces. Thus far, most of the research have developed approaches that are unique to specific domains providing insight into the domain in the process. There is a need for a general statistically robust approach that can be applied to a broad range of domains to uncover variable sequences that lead to outcomes of interest. We propose a computational approach to uncover characteristic response paths of a population from an individual-level multivariate time series data set. The approach is based on a statistical model that accommodates arbitrary and mixed set of distributions for the endogenous variables, accommodates intersession intervals and variables, and reliably estimates individuals' parameters through statistical pooling by uncovering clusters of similar members. These features make the proposed model suitable for a large variety of real-world datasets. We show that using such a model one can extract characteristic paths over possible activity sequences starting from an impulse leading up to a target variable of interest. [ABSTRACT FROM AUTHOR]

We propose an approach to match returning donors to fundraising campaigns on philanthropic crowdfunding platforms. It is based on a structural econometric model of utility-maximizing donors who can derive both altruistic (from the welfare of others) and egoistic (from personal motivations) utilities from donating—a unique feature of philanthropic giving. We estimate our model using a comprehensive data set from DonorsChoose.org—the largest crowdfunding platform for K–12 education. We find that the proposed model more accurately identifies the projects that donors would like to donate to on their return in a future period, and how much they would donate, than popular personalized recommendation approaches in the literature. From the estimated model, we find that primarily egoistic factors motivate over two-thirds of the donations, but, over the course of the fundraising campaign, both motivations play a symbiotic role: egoistic motivations drive the funding in the early stages of a campaign when the viability of the project is still unclear, whereas altruistic motivations help reach the funding goal in the later stages. Finally, we design a recommendation policy using the proposed model to maximize the total funding each week considering the needs of all projects and the heterogeneous budgets and preferences of donors. We estimate that over the last 14 weeks of the data period, such a policy would have raised 2.5% more donation, provided 9% more funding to the projects by allocating them to more viable projects, funded 17% more projects, and provided 15% more utility to the donors from the donations than the current system. Counterintuitively, we find that the policy that maximizes total funding each week leads to higher utility for the donors over time than a policy that maximizes donors' total utility each week. The reason is that the funding-maximizing policy focuses donations on more viable projects, leading to more funded projects, and, ultimately, higher realized donors' utility. This paper was accepted by Kartik Hosanagar, information systems. [ABSTRACT FROM AUTHOR]

Sometimes we desire change, a break from the same, or an opportunity to fulfill different aspects of our needs. Noting that consumers seek variety, several approaches have been developed to diversify items recommended by personalized recommender systems. However, current diversification strategies operate under a one-shot paradigm without considering the evolution of preferences resulting from recent consumption. Therefore, such methods often sacrifice accuracy. In the context of online media, we show that by recognizing that consumption in a session is the result of a sequence of utility-maximizing selections from various categories, one can increase recommendation accuracy by dynamically tailoring the diversity of suggested items to the diversity sought by the consumer. Our approach is based on a multicategory utility model that captures a consumer's preference for different categories of content, how quickly the consumer satiates with one category and wishes to substitute it with another, and how the consumer trades off costly search efforts with selecting from a recommended list to discover new content. Taken together, these three elements allow us to characterize how an individual selects a diverse set of items to consume over the course of a session and how likely the individual is to click on recommended content. We estimate the model using a clickstream data set from a large media outlet and apply it to determine the most relevant content to recommend at different stages of an online session. We find that our approach generates recommendations that are on average about 10% more accurate than optimized alternatives and about 25% more accurate than those diversified using existing diversification strategies. Moreover, the proposed method recommends content with diversity that more closely matches the diversity sought by readers, exhibiting lower concentration–diversification bias than other personalized recommender systems. Using a policy simulation, we estimate that recommending content using the proposed approach would result in visitors reading 23% additional articles at the studied website and deriving 35% higher utility. This could lead to immediate gains in revenue for the publisher and longer-term improvements in customer satisfaction and retention at the site. [ABSTRACT FROM AUTHOR]