Community Detection for Device Graph Impression Deduplication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The decline of third-party cookies and mobile advertising identifiers has led to inaccurate media impression measurement due to the use of alternative identifiers like email and IP addresses, which can misrepresent the true quantity of impressions, especially when devices are shared among multiple users or used in one-off sign-ons, resulting in duplicated and misassigned media impressions.
Innovation Solution
A community detection algorithm is applied to create a device graph using PII-to-device links, allowing for the deduplication of impressions by grouping devices based on interaction frequency, forming communities that represent a single user, and providing accurate device-to-person relationships for demographic assignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If alternative identifiers (email, IP address, Smart TV ID) are used to measure media impressions, then the measurement can continue without third-party cookies, but the accuracy of impression quantity representation deteriorates due to device sharing and one-off sign-ons
Solution Approach 1:
The patent segments the device graph into multiple communities using community detection algorithms, where each community represents a distinct user. This segmentation allows the system to differentiate between multiple users sharing the same device, thereby resolving the overcounting problem and improving measurement precision while maintaining adaptability to alternative identifiers
Solution Approach 2:
The patent introduces device graphs and community detection as an intermediary layer between alternative identifiers and impression measurement. This intermediary structure enables accurate user identification by analyzing interaction patterns and relationships among devices, emails, and users, thus improving measurement accuracy without relying on third-party cookies
2Reliability
If PII-to-device links are used to track impressions, then user identification can be maintained, but duplicate impressions occur when devices are shared among multiple users
Solution Approach 1:
The patent segments the connected device graph into distinct user communities using community detection algorithms. By dividing the graph into separate communities based on interaction patterns, the system can accurately attribute impressions to specific users even when multiple users access the same device, thereby eliminating duplicate impressions while preserving reliable user identification
Solution Approach 2:
The patent changes the parameters used for user identification from simple PII-to-device links to community-based user profiles that consider interaction frequency, device relationships, and behavioral patterns. This parameter transformation enables the system to distinguish between different users on shared devices, reducing duplicate impressions while maintaining identification reliability
3Measurement precision
If community detection is applied to deduplicate impressions, then measurement accuracy improves, but computational complexity increases due to graph processing requirements
Solution Approach 1:
The patent performs preliminary actions by pre-processing the device graph to identify and establish community structures before actual impression deduplication occurs. This advance computation organizes the data into manageable communities, reducing the computational burden during real-time deduplication operations while maintaining high measurement precision
Solution Approach 2:
The patent implements dynamic community detection that can adapt to changing device relationships and user behaviors. The system dynamically updates community assignments based on evolving interaction patterns, allowing efficient processing of large graphs while maintaining accurate deduplication. This dynamic approach reduces computational complexity by only re-processing portions of the graph that have changed
Data Source
AI summary
Methods, apparatus, systems, and articles of manufacture for user identification via community detection are disclosed. Example instructions, when executed, cause at least one processor to at least access personally identifiable information to device links, build a device graph based on the personally identifiable information to device links, split components of the device graph into person clusters using community detection, create a snapshot including a device-to-person link lookup, and prepare a person-level impression measurement report from the snapshot.