Click Value Assessment via Dwell Time Distribution Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Pay Per Performance (PPC) models face challenges in accurately distinguishing between accidental and non-accidental clicks due to arbitrary thresholds and lack of conversion data, leading to unfair charging for advertisers and potential revenue loss for publishers, especially across different syndication networks with varying quality.
Innovation Solution
A data-driven approach that analyzes user interactions by decomposing dwell time distributions into prototypical components to identify measurement thresholds for accidental clicks, using statistical models to estimate parameters and compute reference measurement thresholds for cost adjustments across platforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If arbitrary time thresholds are used to identify accidental clicks, then the system is simple to implement, but the measurement precision of click classification deteriorates
Solution Approach 1:
The patent changes the parameter from a fixed arbitrary time threshold to a dynamically determined threshold based on dwell time distribution analysis. The system analyzes the distribution of user dwell times on landing pages and identifies natural break points in the distribution to determine the threshold, thereby adapting the threshold parameter to actual user behavior patterns rather than using static arbitrary values.
Solution Approach 2:
The patent replaces the simple mechanical threshold-setting approach with a statistical analysis system. Instead of manually setting or using fixed time thresholds, the system uses dwell time distribution analysis and statistical modeling to automatically determine optimal thresholds, substituting data-driven computation for rule-based mechanical decision-making.
2Measurement precision
If conversion data is used to compute discount factors, then the charging accuracy improves, but the system complexity increases due to data availability and quality issues
Solution Approach 1:
The patent introduces dwell time as an intermediary metric that bridges the gap between simple click counting and complex conversion tracking. Instead of directly relying on conversion data which is often unavailable or incomplete, the system uses dwell time—a readily available behavioral signal—as a proxy to infer click quality and compute discount factors, thereby avoiding the complexity of implementing comprehensive conversion tracking while maintaining charging accuracy.
Solution Approach 2:
The patent uses dwell time data, which is easily and cheaply collected from web analytics, as a substitute for expensive and complex conversion data collection. Dwell time is a readily available metric that does not require additional tracking infrastructure or advertiser cooperation, making it a cost-effective alternative for computing click discount factors.
3Productivity
If different bids are used for different syndication networks, then the revenue optimization improves, but the operational complexity increases
Solution Approach 1:
The patent creates a universal bidding framework that works across all syndication networks using a common methodology. Instead of manually setting different bids for each network, the system applies the same dwell time-based discount factor calculation universally across all networks, automatically adapting to each network's characteristics through data-driven analysis while maintaining operational simplicity through a unified approach.
Data Source
AI summary
The present teaching relates to analyzing user behavior associated with web contents. Information related to user interactions associated with a content item placed on a reference property is first obtained. A measurement associated with each user interaction of the content item is determined based on the obtained information. An analyzing model for the content item which characterizes statistics of the measurements associated with the content item is further constructed. A measurement threshold to be used to determine a cost of placing the content item on a target property is further determined using the constructed analyzing model.