with Yu Song, Yiqi Li, and Puneet Manchanda

with Puneet Manchanda and Eric Schwartz (Invited for Revision at Journal of Marketing Research)

A majority of US households view content on online video streaming services, consuming on demand. Not surprisingly, ad spending on such services is growing rapidly. We develop a three-stage approach to deliver an optimal ad schedule that balances the interest of the viewer (content consumption) with that of the streaming platform (ad exposure). In the first stage, we use theoretical findings to develop two parsimonious metrics – Bingeability and Ad Tolerance – to capture the interplay between content consumption and ad exposure. Bingeability represents the number of completely viewed unique episodes of a show while Ad Tolerance represents the willingness of a viewer to continue watching after ad exposure. The second stage uses detailed data on viewing activity and ad delivery to predict these metrics for a viewing session using causal machine learning methods. This is achieved via tree-based algorithms combined with instrumental variables to accommodate the non-randomness in ad delivery. In the third stage, we use the predicted metrics as inputs to a novel constrained optimization procedure that provides the optimal ad schedule. We find that "win-win" ad schedules are possible that allow for a simultaneous increase in content consumption and ad exposure.

Keywords: Advertising Scheduling, Streaming Media, Binge-Watching, Causal Machine Learning, Optimization


with Puneet Manchanda (Job Market Paper)

The influencer marketing industry is growing exponentially because of the increasing popularity of social media stars who primarily reach their audience(s) via custom videos published on a variety of social media platforms (e.g., YouTube, Instagram, Twitter and TikTok). While there has been ample research to study the characteristics of conventional advertising videos and their impact on marketing outcomes, there has been little research on the design and effectiveness of influencer videos. Using publicly available data on YouTube influencer videos, we implement novel interpretable deep learning architectures, supported by transfer learning, to identify significant relationships between advertising content in influencer videos (across text, audio, and images) and video views, interaction rates and sentiment. This is followed up with a second study to investigate whether influencers “learn” these relationships over time. By avoiding ex-ante feature engineering, and instead using ex-post interpretation, our approach avoids making a trade-off between interpretability and predictive ability. We filter out relationships that are affected by confounding factors unassociated with an increase in attention to advertising content, thus facilitating the generation of plausible causal relationships between video design elements and marketing outcomes which can be tested in the field. Our key finding is that brand mentions in the first 30 seconds of a video are on average associated with a significant increase in attention to the brand but a significant decrease in sentiment expressed towards the video. We illustrate the learnings from our approach for both influencers and brands.

Keywords: Video Influencers, Social Media, Interpretable Machine Learning, Deep Learning, Transfer Learning


with Puneet Manchanda

© June 2020 by Prashant Rajaram