Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Distributional Treatment Effects of Content Promotion: Evidence from an ABEMA Field Experiment

Shota Yasui, Tatsushi Oka, Undral Byambadalai, Yuki Oishi

TL;DR

The paper tackles how top-of-screen content promotions shape viewing behavior by estimating distributional treatment effects on total viewing time $Y$, not just mean changes. It leverages a large-scale ABEMA RCT with randomization $D \in \{0,1\}$ and $\rho=0.1$, and defines $\Delta^{DTE}(y)=F_{Y(1)}(y)-F_{Y(0)}(y)$ and $\Delta^{PTE}(y,h)$ to capture shifts across the entire distribution. Methodologically, it uses a regression-adjusted distribution regression framework with pre-treatment covariates, gradient boosting to estimate $F_{Y(d)|X}(y|x)$, and cross-fitting for Neyman-orthogonal inference, validated via $500$ bootstrap replications. Empirically, promotions consistently increase initial engagement, with heterogeneous effects by content type and gender: short-form content often leads to multi-episode viewing, while long-form serialized content yields sustained engagement only when early episodes strong-hook viewers; some cases show trial viewing without lasting retention. The findings inform promotional strategy and content design, guiding targeted allocation of promotions and narrative pacing to maximize retention and cross-episode consumption.

Abstract

We examine the impact of top-of-screen promotions on viewing time at ABEMA, a leading video streaming platform in Japan. To this end, we conduct a large-scale randomized controlled trial. Given the non-standard distribution of user viewing times, we estimate distributional treatment effects. Our estimation results document that spotlighting content through these promotions effectively boosts user engagement across diverse content types. Notably, promoting short content proves most effective in that it not only retains users but also motivates them to watch subsequent episodes.

Distributional Treatment Effects of Content Promotion: Evidence from an ABEMA Field Experiment

TL;DR

The paper tackles how top-of-screen content promotions shape viewing behavior by estimating distributional treatment effects on total viewing time , not just mean changes. It leverages a large-scale ABEMA RCT with randomization and , and defines and to capture shifts across the entire distribution. Methodologically, it uses a regression-adjusted distribution regression framework with pre-treatment covariates, gradient boosting to estimate , and cross-fitting for Neyman-orthogonal inference, validated via bootstrap replications. Empirically, promotions consistently increase initial engagement, with heterogeneous effects by content type and gender: short-form content often leads to multi-episode viewing, while long-form serialized content yields sustained engagement only when early episodes strong-hook viewers; some cases show trial viewing without lasting retention. The findings inform promotional strategy and content design, guiding targeted allocation of promotions and narrative pacing to maximize retention and cross-episode consumption.

Abstract

We examine the impact of top-of-screen promotions on viewing time at ABEMA, a leading video streaming platform in Japan. To this end, we conduct a large-scale randomized controlled trial. Given the non-standard distribution of user viewing times, we estimate distributional treatment effects. Our estimation results document that spotlighting content through these promotions effectively boosts user engagement across diverse content types. Notably, promoting short content proves most effective in that it not only retains users but also motivates them to watch subsequent episodes.
Paper Structure (20 sections, 5 equations, 11 figures, 4 tables)

This paper contains 20 sections, 5 equations, 11 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Experimental Setup at ABEMA
  3. Empirical Framework
  4. Treatment Effects
  5. Estimation Method
  6. Empirical Results
  7. Case 1: Promotion Increases Initial Viewing and First Episode Completion
  8. Case 2: Promotion Encourages Multi-Episode Short-Form Viewing
  9. Case 3: Promotion Strengthens Long-Form Engagement and Completion
  10. Case 4: Promotion Induces Only Trial Viewing Without Sustained Engagement
  11. Key Findings across Content Types
  12. Observed Heterogeneity
  13. Conclusion
  14. Appendix
  15. Pre-treatment Covariates
  16. ...and 5 more sections

Figures (11)

  • Figure 1: Top-of-Screen Promotion at the ABEMA app
  • Figure 2: Illustration of Viewing Time Density across Sequential Episodes
  • Figure 3: Distributional Treatment Effect and Probability Treatment Effect (Case 1)
  • Figure 4: Distributional Treatment Effect and Probability Treatment Effect (Case 2)
  • Figure 5: Distributional Treatment Effect and Probability Treatment Effect (Case 3)
  • ...and 6 more figures