Technology that processes text, audio and video, as well as location data, has revolutionized many industries by enabling innovative operations for customer retention. To retain transactions for a platform and viewers for advertisers, this dissertation leverages novel digital tools to analyze consumer behavior, proposes original economic frameworks to guide platform design, and generates new insights to encourage engaging ad creatives. Chapter 1 and Chapter 2 are devoted to disintermediation, also known as leakage, when users stop doing transactions on the platform. Buyers and sellers can coordinate outside the platform to transact directly, usually to avoid paying fees after being matched. Although platforms are concerned about losing revenue, leakage---by its very nature---is hard to measure and mitigate. Working with Huaiyu Zhu and Jingyi Wang who are employees of an on-demand logistics platform, we develop three detection algorithms and examine their effectiveness in Chapter 1. We find that the state-of-art BERT-based deep learning model improves the precision of detection, but comes with an expensive tradeoff of dropping the high recall we achieve by tracing whether drivers stopped by the origin and destination of a canceled trip. We call for caution regarding heavy resource allocation towards text mining on conversational data when platforms could be more cost-effective by using only behavioral data to detect leakage. In Chapter 2, we focus on the economic incentives behind leakage. We exploit a quasi-experiment that gradually introduced driver commissions, thereby generating variation in participants' incentives for leakage. The introduction of this commission increased leakage by nearly four percentage points, doubling the percentage of offline transactions we detected. We leverage the variation in commission fees to estimate price sensitivities and transaction costs in a structural model. The likelihood of leakage increases as the quoted price of the delivery increases, as the drivers' potential savings in the commission exceed the costs of offline coordination. Our model estimates suggest that customers typically receive half of the commission savings from drivers to rationalize their agreement to leakage. We discuss how targeting, monitoring, and matching can mitigate leakage by better aligning the incentives of different parties in two-sided markets. Chapter 3 (with Joonhyuk Yang, Lakshman Krishnamurthi, and Purushottam Papatla) focuses on ad avoidance when viewers stop watching a video ad. We use a proprietary measure of ad interruption provided by a marketing analytics platform to tracks whether an ad is being played on the screen, and develops a new framework to algorithmically measure the energy level in ad content from the audio of ads. Our machine learning-based measure is related to the human arousal stimulated by ad content. Given that TV ads have become increasingly energetic, we investigate the association between the energy level in ad content and the tendency for consumers to avoid the ads. Overall, more energetic commercials are less likely to be avoided by viewers. However, the association varies across product categories and program genres. We suggest advertisers pay attention to components of ad content other than loudness, which has been regulated by law. Digital transformation is a mindset, not just about the technology solutions. I hope that these three chapters not only document how novel digital measures can be developed to inform and boost retention efforts for platforms and advertisers, but also demonstrate that marketing researchers and economists can seize opportunities in digitization to work with firms closely to create a culture of innovation and customer-centricity.
