Video Streaming QoE Control for Buffering and Engagement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video streaming optimization techniques struggle to balance conflicting metrics such as network bandwidth, throughput, and device limitations, requiring months of manual effort to find optimal parameters, and fail to consider long-term user engagement.
Innovation Solution
A system utilizing counterfactual estimation and Double Machine Learning models to determine causal relationships between video streaming quality and user engagement, optimizing parameters like ABR, compression, and CDN selection for improved short-term and long-term engagement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high-quality video streaming is provided despite poor network bandwidth, then video quality is improved, but buffering frequency increases
Solution Approach 1:
The system dynamically adjusts video quality parameters in real-time based on network conditions and user context. The machine learning model continuously monitors network bandwidth, device capabilities, and user engagement metrics to adaptively select optimal video bitrate and resolution, resolving the contradiction between maintaining high video quality and preventing buffering under varying network conditions.
Solution Approach 2:
The system implements a feedback loop where user engagement metrics (watch time, re-watches, completions) are fed back into the machine learning model to continuously refine quality decisions. This feedback mechanism allows the system to learn from past performance and adjust video quality parameters to balance visual fidelity with streaming reliability, reducing buffering while maintaining acceptable or high quality where possible.
2Loss of time
If delay of playback start is reduced, then user engagement is improved, but frequency of quality-changes during streaming increases
Solution Approach 1:
The system performs preliminary buffering of video content at optimized quality levels before playback begins. The machine learning model predicts initial network conditions and pre-loads appropriate video segments, enabling faster playback start without requiring frequent quality adjustments during streaming. This advance preparation resolves the contradiction by establishing a stable quality baseline early in the streaming session.
Solution Approach 2:
The system strategically changes quality parameters at specific moments in the streaming lifecycle rather than continuously. By concentrating quality decisions at key transition points (initial buffer, segment boundaries) rather than continuously adjusting during playback, the system achieves fast start times while minimizing the frequency of perceptible quality changes, maintaining stability during the streaming portion.
3Measurement precision
If manual parameter tuning is performed to balance conflicting metrics, then optimization accuracy is improved, but time and effort required increases
Solution Approach 1:
The system employs machine learning models that automatically learn and optimize streaming parameters without human intervention. The model trains on historical streaming data and user engagement metrics to autonomously determine optimal quality settings, eliminating the need for manual parameter tuning while achieving superior optimization accuracy through continuous learning from real-world performance data.
Solution Approach 2:
The patent replaces manual mechanical tuning processes with automated machine learning algorithms. Instead of human experts manually adjusting parameters based on A/B testing and analysis, the system uses computational models that automatically process network conditions, device characteristics, and engagement metrics to determine optimal settings, dramatically reducing time and effort while improving precision through data-driven insights.
4Reliability
If current viewing experience is optimized, then immediate user satisfaction is improved, but long-term user engagement may be compromised
Solution Approach 1:
The system takes preliminary actions to protect long-term engagement by monitoring and managing cumulative streaming quality metrics throughout the user session. It proactively adjusts parameters to prevent degradation patterns that could lead to user churn, such as excessive buffering events or quality fluctuations, thereby maintaining both immediate satisfaction and long-term retention through forward-looking quality management.
Solution Approach 2:
The system implements multi-timescale feedback loops that consider both immediate viewing quality and long-term engagement outcomes. User engagement metrics (session duration, return visits, content completions) are fed back into the machine learning model to refine quality decisions over time. This feedback mechanism ensures that optimization strategies balance short-term satisfaction with long-term user retention, adjusting parameters to maintain sustainable quality levels that promote continued engagement.
Data Source
AI summary
Disclosed are systems and methods to optimize the quality of experience of a video streaming session through counterfactual estimation to measure the impact of streaming quality on user engagement. The impact on user engagement may consider both short term user engagement (e.g., current session) and long term user engagement (e.g., user engagement over days, weeks, months, or years).