Cross-Platform Overlap Estimation Using Multivariate Probit Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to accurately determine overlap in reach and impressions across platforms like linear TV, OTT, and YouTube due to lack of device-level identification data for YouTube, leading to under-identified systems of equations.
Innovation Solution
A computer-implemented method using a multivariate probit model to estimate overlap by combining reach statistics from platforms that identify devices with those that do not, through an aggregation server and overlap calculator, employing a multivariate probit model to calculate exposure proportions across multiple platforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional techniques are used to track reach and impressions, then statistics can be obtained for platforms that identify devices (linear TV, OTT), but overlap determination between platforms is impossible when device identification is not provided (YouTube)
Solution Approach 1:
The patent introduces an intermediary modeling system that mediates between platforms with device identification and platforms without device identification. The multivariate probit model acts as a mediator to infer overlap statistics by combining partial information from multiple sources, enabling overlap determination even when direct device-level matching is impossible for certain platforms like YouTube
Solution Approach 2:
The patent transforms the problem from direct device-level matching to a statistical parameter estimation problem. By changing the approach from deterministic identification to probabilistic modeling with multivariate probit functions, the system can estimate overlap parameters (reach, frequency, impressions) without requiring direct device identification from all platforms
2Measurement precision
If device-level identification data is collected from all platforms, then accurate overlap determination is possible, but system complexity and data processing requirements increase significantly
Solution Approach 1:
The patent applies partial action by collecting device identification data only from platforms where it is available (linear TV, OTT) rather than requiring it from all platforms. The modeling system compensates for the missing data from platforms like YouTube through statistical inference, achieving sufficient measurement precision without the excessive complexity of universal device-level tracking
Solution Approach 2:
The patent segments the measurement system into two distinct components: (1) platforms with device identification capability that provide direct measurement data, and (2) platforms without device identification that provide aggregate statistics. The multivariate probit model integrates these segmented data sources, allowing accurate overlap determination without requiring uniform device-level data collection across all platforms
Data Source
AI summary
A multivariate probit model is used to determine overlaps for reach and impressions for a plurality of different platforms.