Online System Household Reach Inference via De-biased ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Online systems face challenges in generating accurate analytics for content items presented by multiple content publishers due to inconsistent tracking of events and lack of household-specific information, limiting the ability to tailor content effectively to target audiences across different platforms.
Innovation Solution
An online system employs tracking mechanisms and interactive elements to collect presentation data, identifies users through user-identifying information, retrieves household attributes, and uses machine-learning models to infer the reach of content items for each household, adjusting for population differences through de-biasing processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If content items are presented by multiple content publishers, then the reach and visibility of content items are improved, but the accuracy of analytics becomes worse due to inconsistent tracking
Solution Approach 1:
The patent introduces an intermediary system (the online system) that acts as a mediator between content publishers and the analytics aggregation process. This intermediary collects presentation data from multiple publishers, normalizes it against household profiles, and generates unified analytics, thereby resolving the tracking inconsistency problem while maintaining broad reach across publishers.
Solution Approach 2:
The system creates a universal analytics generation approach that works across multiple content publishers by using a common household profile framework. This universal approach allows the system to aggregate and analyze data from any number of publishers consistently, maintaining measurement precision while expanding reach.
2Adaptability or versatility
If content items are presented by multiple content publishers, then the audience coverage is improved, but the reliability of household-specific analytics becomes worse due to lack of consistent household information
Solution Approach 1:
The patent implements preliminary action by pre-building comprehensive household profiles that contain standardized demographic and behavioral information before content presentation occurs. These pre-established profiles serve as reliable reference data when analytics need to be generated for multiple publishers, ensuring consistent and reliable household-specific analytics across the entire audience coverage.
Solution Approach 2:
The system changes the parameters for household identification and data collection across different content publishers by using a standardized set of household profile attributes. This parameter standardization ensures that household-specific analytics remain reliable and comparable across all publishers while maintaining broad audience coverage.
3Measurement precision
If the online system collects detailed tracking data from all content publishers, then the accuracy of analytics is improved, but the device complexity and data processing requirements increase
Solution Approach 1:
The patent extracts and isolates the complex data processing and analytics generation functions into a separate, centralized system component. By taking out these complex operations from the content publishers and consolidating them in the online system, the solution maintains high measurement precision while reducing the complexity burden on individual publishers and their tracking infrastructure.
Data Source
AI summary
An online system, such as an online content distribution system, receives information describing presentations of a content item. The online system identifies online system users included among individuals presented with the content item, retrieves information describing a household of each user, and de-biases this information to adjust for differences between the population of online system users and the individuals to whom the content item was presented. The online system then trains a set of machine-learning models to infer a reach of the content item for each of multiple households using the de-biased information. Using the machine-learning models, the online system infers a reach of the content item for each of the households.