The dissertation consists of three separate essays that lie at the interface of social network analytics and Customer Relationship Management (CRM). Essay 1 and Essay 2 cover completed research, while the research covered in Essay 3 is at a more preliminary stage.', 'Essay 1: Starting Cold: The Power of Social Networks in Predicting Non-Contractual Customer Behavior', "In this work, we provide an integrated framework for marketing managers on how to appropriately measure and manage customer behavior in a non-contractual setting in the presence of social network data. Customer behavior is directly tied to customer lifetime value (CLV) and customer equity (CE). Predicting customer behavior and their spending patterns, and consequently CLV, in such settings is a very challenging problem due to the absence of a formal declaration of termination of the customer-firm relationship. This implies that inactivity does not necessarily signal the end of the relationship, as a user may temporarily become dormant, and return at a later point in time. Distinguishing between dormant and churned consumers is a hurdle for marketeers who need to allocate their limited resources in a way that increases the overall value of a business’s customer base. Another important implication of non-contractual relationships is evident in customer-based corporate valuations (CBCV). Performing a CBCV requires knowing ahead of time how long a customer will remain with the firm, which inevitably makes non-contractual businesses prone to misvaluations. Therefore, any improvement in the ability to predict behavior in non-contractual settings is highly valuable. In this work, we study the extent to which knowledge of a customer’s social network can enhance the accuracy of forecasting their behavior in terms of future: (1) activity, (2) transaction levels and (3) membership to the group of best customers. We conduct a dynamic analysis on a sample of approximately one million users from the most popular peer-to-peer (P2P) payment application, Venmo. Our models produce high quality forecasts and demonstrate that social networks lead to a significant boost in predictive performance primarily during the first month of a customer's lifetime, thus providing a remedy to the “cold-start†problem. Finally, we characterize significant structural differences with regard to network centrality, density and connectivity between the top 10% and bottom 90% of users immediately after joining the service. We discuss how these structural dissimilarities provide a path towards proactive marketing and improved customer acquisition efforts.", 'Essay 2: Finding Strong Ties in a Facebook Haystack: A Multilayer Social Network Approach', 'In this work, we investigate the question of identifying the strong ties of an individual by just inspecting the person’s underlying social network structure. Strong ties have been documented to play an influential role in people’s decision making process across various settings. From our decision to donate goods to our decision to turn up and vote at the elections, strong ties are the ones who exert the greatest influence on us as they convey greater trust. The digital age has re-emphasized the importance and complexity of this task, as more and more companies have now access to online friendship data of their customers. We use and extend the "social bow tie" framework introduced in \\cite{bowtie} and apply it to a unique dataset from Venmo, the most popular P2P mobile payment application, to expand our knowledge on tie strength prediction. Our dataset is unique because it combines two different but overlapping social networks. On the one hand, we have the Venmo social graph, which comprises of all friend relationships of users that signed up with Facebook (FB). On the other hand, we have the Venmo transactional graph which reflects offline transactions among users. By following the money trail, we are able to differentiate with whom a user is really closely connected to among his FB friends, and we study the extent to which knowledge of a customer’s egocentric FB social network can enhance the accuracy of forecasting whether two individuals: (1) will transact at least once, (2) whether this transaction will be reciprocated and (3) their total number of transactions. Our models produce high quality forecasts for the tasks of predicting the formation of a financial relationship and its reciprocity, yielding final Accuracy scores in the range of 43%-90% and Area Under the Precision-Recall Curve (AUPRC) values in the range of 85%-98%, depending on the exact problem formulation. For the task of predicting the total number of transactions between a pair of users, we get a Mean Square Error (MSE) in the range of 7.38-25.48 and an R-squared in the range of 0.24-0.58. The most informative predictors are found to be the overlap of friends between two individuals, and the clustering coefficient of their non-overlapping friends. These findings are consistent with 1) Granovetter\'s hypothesis: the stronger the tie between any two individuals, the higher the fraction of friends they share in common, and 2) Bott\'s hypothesis: the higher the degree of clustering in an individual’s network the less likely to form a tie with somebody outside the group.', 'Essay 3: Venmo for Change: The Effect of Digital Donations on Customer Engagement', "In today's competitive and connected environment, organizations are investing in corporate social responsibility (CSR) activities to differentiate themselves and create a meaningful engagement with their customers. Digital platforms have reemphasized this need by introducing new forms of donating mechanisms that use social cues to inform the users about a fundraising event. Research has documented the benefits of CSR activities to organizations in terms of enhanced consumer perceptions of the company, but there is little empirical evidence on the effect of digital platform donations on customer engagement as this is expressed by any potential interaction two existing users might have on the platform. In this work, we propose a setting to empirically explore this question. Specifically, we use data from charitable fundraising events in Venmo to investigate whether two users that have contributed to the same charity event and have not previously transacted up to that point in time are more likely to transact after the charity event. Our charitable events are created by exogenous random shocks (e.g., physical catastrophes), which allow us to causally identify the effect of donations on customer engagement. We seek to test whether donating to a common cause increases the likelihood of forming a relationship between two users and whether this likelihood is a decreasing function of the shortest path distance between the two users."]

Last modified

• 10/28/2019

Creator

• Panteleimon Loupos

DOI

• https://doi.org/10.21985/n2-97cf-xn53

Subject

• Operations Management

Keyword

• Cold-Start
• Predictive Analytics
• Peer-to-Peer Financial Networks
• Social Networks
• Customer Behavior
• CRM

Date created

• 2018-01-01

Resource type

• Dissertation

Rights statement

• In Copyright