Inference Engine for Reach Analytics
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 user-specific information, leading to limited ability to tailor content effectively across different platforms.
Innovation Solution
An online system employs tracking mechanisms and machine-learning models to infer reach for content items across multiple publishers by de-biasing impression frequencies and adjusting inferred reaches using linear or exponential decay approaches, ensuring accurate analytics without relying on publishers' communication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If content publishers track events and communicate information to the online system, then analytics accuracy improves, but device complexity and implementation difficulty increase for publishers
Solution Approach 1:
The online system performs self-service by autonomously inferring reach and impression frequency analytics using its own user profiles and content delivery records, without requiring publishers to implement tracking code or communicate event data. The system queries its internal databases to determine which users received content from which publishers and calculates analytics independently.
Solution Approach 2:
The online system acts as an intermediary that bridges the gap between content publishers and analytics needs. Instead of relying on publishers to provide tracking data, the online system mediates by using its own user profile information and content delivery records to infer the required analytics, thereby eliminating the need for publisher-side tracking implementation.
2Measurement precision
If the online system requests user-specific information from content publishers, then analytics accuracy improves, but loss of time and increased device complexity occur
Solution Approach 1:
The online system performs preliminary action by maintaining user profile information and content delivery records in advance. When analytics are needed, the system queries its pre-existing internal databases rather than collecting data in real-time from publishers, significantly reducing data collection time while maintaining accuracy.
Solution Approach 2:
The online system uses its own pre-maintained user profile information and content delivery records to generate analytics independently, eliminating the time-consuming process of requesting and collecting data from multiple publishers. The system serves its own analytics needs using internally available data.
3Measurement precision
If content publishers maintain detailed user-specific information, then analytics accuracy improves, but device complexity and data storage requirements increase
Solution Approach 1:
The online system merges the analytics generation function into its existing user profile and content delivery infrastructure. Instead of requiring publishers to maintain separate tracking systems, the online system combines its user profile database with content delivery records to infer analytics, consolidating data storage and processing requirements into a single centralized system.
Solution Approach 2:
The online system uses its own user profile information to determine impression frequency and reach analytics independently, eliminating the need for publishers to maintain detailed user-specific information. The system serves its own analytics needs by querying its existing user database and content delivery records.
4Adaptability or versatility
If the online system provides analytics for content items presented by multiple publishers, then content-providing user engagement improves, but measurement precision decreases due to inconsistent tracking
Solution Approach 1:
The online system performs self-service by autonomously inferring reach and impression frequency analytics using its own user profiles and content delivery records, without relying on publishers to provide tracking data. This approach maintains measurement precision while enabling multi-publisher analytics capability.
Solution Approach 2:
Instead of the conventional approach where publishers provide tracking data to the online system, this invention inverts the process by having the online system independently infer analytics from its own internal data. This reversal eliminates the accuracy problems caused by inconsistent publisher tracking while maintaining the ability to analyze content performance across multiple publishers.
Data Source
AI summary
An online system receives information describing presentations of a content item to individuals. The online system determines impression frequencies for online system users included among the individuals, de-biases this information, and trains a set of machine-learning models to infer a reach for each of multiple impression frequencies using the de-biased information. The online system predicts the reach for each impression frequency using the models, determines an inferred total number of presentations of the content item based on the inferences, and compares the inferred total number of presentations to a known total number of presentations of the content item. If the inferred total number of presentations is greater than the known total number of presentations, the online system adjusts the inferred reach for a set of impression frequencies by reducing the inferred reach for a highest impression frequency and by increasing the inferred reach for one or more lower impression frequencies.