Real-Time Attribution Modeling Using User-Specified Segments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional attribution systems are inflexible and inefficient, as they require preconfiguration of databases for specific attribution models and struggle to adapt to changing user requests, limiting their ability to generate various reports and perform deeper analyses.
Innovation Solution
A system that stores raw, unprocessed analytics data in a database with multiple nodes, allowing for flexible and efficient application of user-specified attribution models at query time, enabling real-time attribution segmentation and comparison across different models and segments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional systems preconfigure databases for specific attribution models, then they can efficiently generate reports for those models, but they cannot adapt to new models or segments without reconfiguration
Solution Approach 1:
The system implements a universal database schema that can accommodate multiple attribution models (first touch, last touch, linear, time decay, etc.) and various segmentation types (device, geography, demographics, custom) without requiring separate configurations. The database is designed to store raw touchpoint data in a standardized format that can be queried using different models on demand, making the system multi-functional and adaptable to changing analytical needs.
Solution Approach 2:
The system employs dynamic query-time processing where attribution models are applied at the time of data retrieval rather than being statically pre-configured. This allows the system to flexibly switch between different attribution models and segments without reconfiguration, as the modeling logic is parameterized and can be adjusted dynamically based on user requests.
2Adaptability or versatility
If conventional systems reconfigure databases to accommodate new attribution models, then they can support new models, but they require additional computing resources and processing time
Solution Approach 1:
The system performs preliminary data collection and standardization by storing all raw touchpoint data in a unified format during data ingestion. This preliminary action ensures that when attribution models need to be applied, the data is already prepared and can be processed efficiently without requiring time-consuming reconfiguration or transformation operations.
Solution Approach 2:
The system uses parameterized query templates that can be instantiated with different attribution model parameters without copying or duplicating the underlying data storage structure. This allows new attribution models to be supported by simply configuring query parameters rather than creating new database schemas or processing pipelines, maintaining high processing efficiency.
3Speed
If conventional systems pre-process data before administrator requests, then they can quickly generate reports, but they are incapable of generating other attribution reports
Solution Approach 1:
The system implements dynamic data processing where the degree and type of data processing are determined at query time based on the specific attribution model and segmentation requirements. Rather than statically pre-processing data for a single model, the system can adaptively process data on-demand, balancing speed and versatility by caching frequently accessed aggregated data while maintaining the ability to perform detailed processing when needed.
Solution Approach 2:
The system applies different levels of data processing and aggregation to different data subsets based on local requirements. Frequently accessed segments and popular attribution models benefit from pre-computed aggregations for fast retrieval, while less common queries trigger more detailed on-demand processing, optimizing the balance between speed and adaptability for different parts of the system.
Data Source
AI summary
The present disclosure relates to performing attribution modeling in real time using user-specified segments of touchpoint data retrieved from a database using a user-specified attribution model. For example, in one or more embodiments, a system stores raw touchpoint data in a database comprising an aggregator and a plurality of nodes. In particular, each node stores touchpoint data associated with a different user. Upon receiving a first query, the system can, in real time, generate and provide a first digital attribution report based on the stored touchpoint data. Upon receiving a second query, the system can generate a second digital attribution report for a user-specified segment of the touchpoint data represented in the first digital attribution report. Specifically, the system retrieves touchpoint data associated with the user-specified segment from the nodes of the database and uses the aggregator to combine the data to generate the second digital attribution report.