Video Engagement Prediction for Ad Insertion Timing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video sharing platforms face challenges in improving viewer experience due to the lack of identifying markers in video content, making it difficult to determine appropriate times for inserting recommendations or ads without distracting viewers, especially with the vast number of new videos being uploaded daily.
Innovation Solution
A method that predicts user engagement by determining engagement metrics and characterizing features of watched videos, establishing a relationship between these features and metrics, and applying this knowledge to unwatched videos to forecast engagement levels, allowing for enhanced user experiences through targeted content insertion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If content such as recommendations or ads are inserted into watched videos, then viewer experience can be improved, but if inserted at wrong times, the content becomes distracting and harms the viewer experience
Solution Approach 1:
The system performs preliminary analysis of video content to identify engaging and non-engaging portions before inserting recommendations or ads. By pre-determining optimal insertion points based on engagement predictions, the system avoids distracting viewers during high-engagement segments while maximizing the effectiveness of inserted content.
Solution Approach 2:
The system uses engagement metrics and viewer behavior feedback to dynamically adjust where and how content is inserted. By continuously monitoring viewer engagement levels and using this feedback to guide insertion decisions, the system optimizes the balance between improving viewer experience and avoiding distraction.
2Ease of operation
If content is inserted to improve viewer experience, then engagement can be enhanced, but video content contains very few identifying markers making it difficult to determine appropriate insertion points
Solution Approach 1:
The system introduces engagement predictions as an intermediary metric to bridge the gap between video content and insertion decisions. Instead of relying on traditional identifying markers that are scarce in video content, the system uses predicted engagement levels as a mediator to determine optimal insertion points for recommendations and ads.
Solution Approach 2:
The system transforms the problem from detecting identifying markers to measuring engagement metrics. By changing the parameter from structural video features to behavioral engagement data, the system gains abundant information for determining insertion points, as engagement metrics provide continuous feedback throughout the video.
3Productivity
If engagement predictions are made for all videos, then content insertion can be optimized, but the vast number of new videos uploaded daily makes this challenging
Solution Approach 1:
The system segments the video processing task by analyzing videos in portions rather than as complete units. This allows engagement predictions to be made incrementally as video portions are processed, enabling scalable handling of large volumes of uploaded videos while maintaining optimization quality.
Data Source
AI summary
User engagement in unwatched videos is predicted by collecting and aggregating data describing user engagement with watched videos. The data are normalized to reduce the influence of factors other than the content of the videos on user engagement. Engagement metrics are calculated for segments of watched videos that indicate user engagement with each segment relative to overall user engagement with the watched videos. Features of the watched videos within time windows are characterized, and a function is learned that relates the features of the videos within the time windows to the engagement metrics for the time windows. The features of a time window of an unwatched video are characterized, and the learned function is applied to the features to predict user engagement to the time window of the unwatched video. The unwatched video can be enhanced based on the predicted user engagement.