Query-Time Attribution System for Arbitrary Analytics Parameters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional attribution systems are inflexible and inefficient, as they rigidly collect data for preset attribution models and distribution channels, struggle to adapt to changing requests, and require significant computing resources, making real-time query-time attribution modeling unfeasible due to the need for preconfiguration and node cross-talk in distributed architectures.
Innovation Solution
A query-time attribution system that stores raw, unprocessed analytics data in a database with multiple nodes, each storing touchpoint data for a user, allowing for real-time processing and generation of digital attribution reports for arbitrary analytics parameters specified by administrators, eliminating the need for preconfiguration and reducing node cross-talk by storing data per user.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional systems preconfigure databases to store touchpoint data for preset attribution models, then the systems can perform attribution analysis for those specific models, but the systems become inflexible and unable to model touchpoint data for other parameters
Solution Approach 1:
The system implements a universal attribution model that can analyze touchpoint data for any parameter or metric without requiring preconfiguration. The database schema and processing logic are designed to handle arbitrary analytics parameters, allowing the same infrastructure to serve multiple analytical purposes dynamically based on query requirements
2Adaptability or versatility
If conventional systems reconfigure the database to accommodate new attribution requests, then the systems can adapt to changing administrator requests, but additional computing resources and processing time are required
Solution Approach 1:
The system performs preliminary setup by establishing a universal, parameter-agnostic database structure and processing framework in advance. This preconfiguration enables the system to handle any specific attribution request without requiring subsequent reconfiguration, as the infrastructure is already designed to accommodate diverse analytical parameters from the outset
3Quantity of substance
If conventional systems store billions of pieces of data in distributed architecture, then the systems can handle large-scale analytics, but the time required to run query-time attribution analysis becomes substantial and prohibitive
Solution Approach 1:
The system segments the distributed database architecture by organizing data around user identifiers rather than distributing across multiple nodes requiring cross-talk. Each user's touchpoint data is stored together in a single location, eliminating the need for complex distributed queries and node coordination, thereby dramatically reducing query processing time while maintaining the ability to handle billions of records
Data Source
AI summary
The present disclosure relates to performing attribution modeling in real time using touchpoint data that correspond to arbitrary analytics parameters (e.g., a user-specified dimension) and are retrieved from a database using an attribution model. For example, in one or more embodiments, a system stores raw data in an analytics database that comprises an aggregator and a plurality of nodes. In particular, each node stores touchpoint data associated with a different user. Upon receiving a query, the system can, in real time, retrieve subsets of the touchpoint data that correspond to a user-specified dimension in accordance with an attribution model. The system then combines the subsets of touchpoint data using the aggregator and generates the digital attribution report using the combined data.