Audience Recognition Model for Ad Frequency Caps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advertisement delivery systems face challenges in determining advertisement frequency for households with multiple viewers, as they struggle to identify individual viewers and manage cross-channel frequency caps, especially when using multiple mediums such as OTT and traditional television.
Innovation Solution
An advertisement delivery system that utilizes real-time audience recognition data, user activity data, and demographic information to estimate the likelihood of advertisement exposure at a user level, generating a statistical distribution model to enforce cross-channel frequency caps at a household level without requiring deterministic viewer identities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If advertisement delivery systems rely on impressions to determine advertisement frequency, then advertisement presentation control is simplified, but the system cannot determine whether advertisements were presented to particular viewers in households with multiple viewers
Solution Approach 1:
The patent introduces an audience recognition model as an intermediary component that processes household-level impression data and viewer activity data to infer individual viewer exposure. This model acts as a mediator between the simple impression tracking system and the need for precise viewer-level frequency measurement, enabling the system to determine which specific viewers were exposed to advertisements without fundamentally redesigning the entire advertisement delivery infrastructure.
Solution Approach 2:
The patent replaces the traditional mechanical approach of direct viewer identification (requiring individual user accounts, device tracking, and deterministic identification) with a statistical inference system. The audience recognition model uses probabilistic methods to substitute for the impossible deterministic identification, allowing the system to estimate viewer exposure based on available household-level data and viewer activity patterns.
2Reliability
If the system cannot determine individual viewer exposure, then frequency cap control at viewer level becomes impossible, but implementing household-level tracking increases system complexity
Solution Approach 1:
The patent segments the frequency cap control problem into two distinct levels: household-level impression tracking and individual viewer inference. The system maintains separate data structures for household impressions and viewer activity, processing them through different components (impression log processor and audience recognition model). This segmentation allows the system to enforce frequency caps reliably at the viewer level while keeping the overall system architecture manageable by dividing complex processing into specialized modules.
Solution Approach 2:
The patent adds a new dimension to the data processing architecture by introducing a probabilistic inference layer that operates parallel to the traditional impression tracking system. Rather than simply extending the existing linear tracking chain, the system creates a two-dimensional processing structure where deterministic impression data coexists with probabilistic viewer exposure estimates, enabling frequency cap control without collapsing system complexity.
3Adaptability or versatility
If the system lacks ability to determine combined advertisement frequency across multiple mediums, then cross-channel frequency management is impossible, but integrating multiple data sources increases measurement complexity
Solution Approach 1:
The patent creates a universal audience recognition model that can process multiple types of input data from different mediums (OTT impressions, television impressions, web browser data, third-party platform data) through a single unified processing framework. This multi-functional model maintains consistent probabilistic inference methods across diverse data sources, enabling the system to manage frequency caps uniformly across all channels while preserving measurement accuracy through standardized processing procedures.
Solution Approach 2:
The patent combines multiple heterogeneous data sources (structured impression logs, unstructured viewer activity data, demographic information, contextual data) into a composite information structure that feeds the audience recognition model. This composite approach integrates diverse data types with different characteristics and quality levels, creating a unified probabilistic estimate of viewer exposure that leverages the strengths of each data source while compensating for individual limitations through their combination.
Data Source
AI summary
Devices, systems, and methods are provided for audience recognition. A method may include receiving over-the-top (OTT) advertisement impression data comprising metadata and content of advertisement bid requests, the metadata indicative of scheduled OTT media presentation; receiving user activity data indicative of day-part times when viewers watch content absent from the OTT advertisement impression data; generating, based on the OTT advertisement impression data, a first demographic probability vector; generating, based on the user activity data, a second demographic probability vector; generating, based on a combination of the first demographic probability vector and the second demographic probability vector, a third demographic probability vector, each entry of the third demographic probability vector indicative of a third probability that a viewer is in a respective age range; and generating an indication of the third demographic probability vector.