Analytics Report Generation with Pseudo-Users for Unidentified Events
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Website owners face challenges in understanding user interactions and making data-driven decisions due to gaps in information, especially when users interact with different browsers and devices, and offline purchases, leading to incomplete analytics and decision-making obstacles.
Innovation Solution
Utilizing machine-learned prediction models to estimate business metrics from unidentified events, combining them with identified data to generate comprehensive analytics reports, including pseudo users and sessions, and adjusting processing parameters based on generated data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional analytics methods are used that rely on stable user identifiers, then data accuracy for identified users is maintained, but data completeness deteriorates due to gaps from unidentified events and cross-device tracking limitations
Solution Approach 1:
The patent introduces machine-learned prediction models as an intermediary between identified and unidentified events. These models predict user characteristics, session metrics, and conversion probabilities for unidentified events, effectively mediating the gap between complete but imprecise identified data and incomplete but potentially accurate unidentified data.
Solution Approach 2:
The system creates synthetic copies of user data through pseudo-users for unidentified events. By generating predicted user profiles, session data, and conversion metrics for unidentified events, the system replicates the structure and characteristics of identified user data, allowing comprehensive analytics without requiring actual user identifiers.
2Loss of information
If machine-learned prediction models are used to estimate metrics from unidentified events, then data completeness improves, but computational resources and processing time increase
Solution Approach 1:
The system applies machine-learned prediction models selectively rather than uniformly to all events. Prediction models are primarily applied to unidentified events where they add value, while identified events continue to be processed through traditional methods, avoiding unnecessary computational overhead.
Solution Approach 2:
The system pre-trains machine-learned prediction models using historical identified event data before deployment. This preliminary training phase allows the models to learn patterns and relationships offline, reducing the computational burden during real-time processing of unidentified events.
3Reliability
If comprehensive analytics data is generated including pseudo-users and predicted metrics, then decision-making quality improves, but system complexity increases
Solution Approach 1:
The system segments analytics processing into distinct pathways: one for identified events using traditional methods and another for unidentified events using machine-learned prediction. This segmentation allows each pathway to be optimized independently and simplifies the overall system architecture by clearly defining boundaries between different processing modes.
Solution Approach 2:
The machine-learned prediction models serve multiple functions simultaneously: they predict user characteristics, estimate session metrics, calculate conversion probabilities, and generate pseudo-user profiles. This multi-functionality reduces system complexity by consolidating multiple prediction tasks into a single unified model framework.
Data Source
AI summary
Techniques for generating a report for a website are presented herein. A method can include accessing, by one or more computing devices, a plurality of unidentified events. Each event in the plurality of unidentified events can be associated with one or more properties. Additionally, the method can calculate, using a machine-learned prediction model, a number of pseudo users associated with the plurality of unidentified events based on an event-to-user-ratio and a total number of unidentified events. Moreover, the method can include assigning a first event from the plurality of unidentified events to a first pseudo user based on the one or more properties of the first event. Furthermore, the method can include generating the report for the website. The report includes information derived from the first event being assigned to the first pseudo user.