Feed Layout Optimization via Scrolling Behavior Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Online publishers face the challenge of users navigating away from feeds after viewing only a few content items, as existing greedy algorithms prioritize content with higher click-through rates without considering user scrolling behavior and display properties.
Innovation Solution
A behavioral modeling system is used to optimize content placement in online feeds by calculating a predicted continuous scrolling probability, taking into account click-through rate, content item height, distance from the top, and screen size, to ensure users view more content items, with content items being ordered to maximize the probability of continued scrolling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a greedy algorithm is used to place content items with higher click-through rate near the top of the feed, then the click-through rate is improved, but the number of content items viewed by users decreases
Solution Approach 1:
The patent changes the optimization parameter from immediate click-through rate to predicted continuous scrolling probability. The system calculates this probability by modeling user scrolling behavior and considering multiple factors including content item attributes (height, distance from top) and user state attributes (screen size, current position in feed). This parameter transformation resolves the contradiction by prioritizing content that keeps users scrolling rather than content that immediately attracts clicks.
Solution Approach 2:
The patent implements feedback through continuous modeling and updating of user scrolling behavior. The system observes actual user scrolling patterns and uses this feedback to refine the behavioral model, which then improves the prediction of continuous scrolling probability. This feedback loop allows the system to adapt to changing user preferences and resolve the contradiction between immediate engagement and continued viewing.
2Productivity
If content items are placed to maximize immediate click-through rate, then user engagement at the top of the feed is improved, but user retention and continued viewing decreases
Solution Approach 1:
The patent applies preliminary action by pre-calculating the predicted continuous scrolling probability for each content item before it is displayed. The system models user behavior and predicts how users will interact with the feed sequence in advance, allowing content items to be ordered to maximize both immediate engagement and continued viewing. This preliminary modeling resolves the contradiction by planning the feed sequence to maintain user interest throughout.
Solution Approach 2:
The patent introduces dynamics by making the feed ordering adaptive rather than static. The system continuously updates the behavioral model based on actual user scrolling behavior and uses this dynamic information to reorder content items. This dynamic adjustment allows the feed to optimize both immediate engagement and user retention in real-time, resolving the contradiction between initial engagement and continued viewing.
Data Source
AI summary
A method and apparatus for populating content items into a feed is provided. The feed comprises a sequence of content item ordered in such a way as to maximize a number of content items displayed to a user by virtue of the user scrolling down through the feed. The content items are each associated with a click-through rate, an indication of a number of times the content has been displayed to users, an indication of a number of times that the users have scrolled to a next item in the feed after the item was displayed, and a height of the content item. These values are used to train a behavioral model and then used by the behavioral model to layout the content items in a feed rendered at a user device.