Probabilistic Viewership Assignment via Markov Chain
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods face challenges in accurately projecting person-level viewership from household-level data, as they lack information on individual viewers, leading to inaccuracies and large margins of error due to small sample sizes of person-level data.
Innovation Solution
A probabilistic minute-by-minute assignment model using a Markov Chain is employed, which accesses panelist and household data to determine the probability of individual members watching a program by analyzing total watched minutes and continuous series of watched states, integrating person-level and household-level data for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If household-level tuning data is used to project viewership, then data coverage is improved, but measurement precision deteriorates due to lack of individual viewer information
Solution Approach 1:
The patent introduces a probabilistic assignment model as an intermediary mechanism that connects household-level tuning data with person-level viewership estimation. The model uses Markov Chain methodology to assign probabilities of program watching to individual household members based on available household data, thereby enabling person-level measurement precision while maintaining household-level data coverage.
Solution Approach 2:
The patent transforms the measurement approach by changing from deterministic assignment to probabilistic assignment. By introducing probability parameters and using Markov Chain state transitions, the system can estimate individual viewership likelihoods from aggregate household data, resolving the contradiction between data coverage and measurement precision.
2Measurement precision
If person-level panelist data is used, then measurement precision is improved, but reliability deteriorates due to small sample sizes
Solution Approach 1:
The patent merges household-level tuning data with person-level panelist data through the probabilistic assignment model. By combining these two data sources, the system achieves both the measurement precision of person-level data and the statistical reliability of large-scale household data, eliminating the need to choose between the two.
Solution Approach 2:
The patent segments the viewership measurement problem into household-level tuning events and individual member probability assignments. This segmentation allows the system to process large volumes of household data while generating person-level insights, thereby improving both reliability through large samples and precision through individualized probability estimates.
3Device complexity
If traditional aggregation methods are used, then device complexity is reduced, but information loss increases due to inability to track individual viewing behavior
Solution Approach 1:
The patent introduces dynamic probabilistic assignment that adapts to individual household compositions and viewing patterns. Rather than using static aggregation, the system dynamically calculates watching probabilities for each household member based on household-specific tuning data, preserving individual viewing information while maintaining manageable system complexity through modular Markov Chain implementation.
Data Source
AI summary
Techniques for projecting person-level viewership from household-level tuning events are described. Initially, panelist viewing data are accessed and a plurality of state values based on the panelist viewing data are determined. Then, tuning data representing tuning events associated with particular households are accessed. For at least one tuning event represented by the tuning data, household member data is accessed, a portion of the panelist viewing data whose panelist information matches at least a portion of the member data is determined, a total number of watched minutes of the program by an individual member and a number of continuous series of watched states of the program by the individual member is determined, and an output representative of a probability that the particular portion of the program was watched by one or more of the individual members is generated.