Credit Portfolio Benchmarking via Pre-Integrated Data Files
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automated tools for comparing credit portfolios among lending institutions are inadequate, failing to provide effective benchmarking information that allows institutions to assess their portfolio performance relative to peers.
Innovation Solution
A credit portfolio benchmarking system that queries massive amounts of current and historical trade data, combining consumer and trade data to generate reports based on user-defined criteria, including population definition, segmentation, and attribute selection, without requiring a query-time join of trade and consumer data, and provides tiered service levels for different levels of information disclosure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional automated tools are used for portfolio comparison, then the system complexity is low, but the benchmarking information quality and accuracy are inadequate
Solution Approach 1:
The system segments the benchmarking process into distinct functional modules: data acquisition module that collects trade and consumer data from multiple sources, data processing module that integrates and cleanses the data, analysis module that performs portfolio comparisons, and reporting module that generates benchmarking reports. This segmentation allows each module to be optimized independently for accuracy while managing overall system complexity.
Solution Approach 2:
The system introduces an intermediary data processing layer that sits between raw data sources and the analysis functions. This intermediary layer includes data validation rules, integration protocols, and quality assurance mechanisms that enhance benchmarking accuracy without requiring changes to the underlying data sources or analysis algorithms.
2Productivity
If trade data and consumer data are joined at query run time, then the data integration is flexible, but the query execution time increases significantly
Solution Approach 1:
The system performs data joins and integrations in advance during an offline data preparation phase, creating pre-integrated data structures before query execution. This preliminary action includes joining trade data with consumer data, aggregating portfolio metrics, and preparing benchmarking datasets, so that online queries can execute rapidly without performing complex joins at query run time.
Solution Approach 2:
The system extracts and separates the data integration function from the query execution function. The integration process is extracted to an independent offline processing step, while the online query system only needs to retrieve and analyze pre-integrated data. This separation eliminates the time penalty of runtime joins while maintaining data flexibility.
3Loss of information
If detailed peer information is disclosed in benchmarking reports, then the information utility is high, but confidential information of specific peers is revealed
Solution Approach 1:
The system applies different levels of information disclosure to different portions of the benchmarking report based on local requirements. Aggregate portfolio metrics and industry comparisons are disclosed in full to maintain information utility, while identifiers and specific confidential details of individual peer institutions are masked or suppressed to protect confidentiality. This local quality approach allows simultaneous achievement of transparency and privacy protection.
Data Source
AI summary
A portfolio benchmarking system comprises a repository of trade data, a repository of consumer data, a build computer, and a benchmarking query application. The repository of trade data comprises a plurality of data items regarding trade lines. The repository of consumer data comprises a plurality of data items regarding consumers, wherein at least some information in the consumer data is not in the trade data and at least some information in the trade data is not in the consumer data. The build computer periodically generates at least one data file comprising a plurality of data items, each data item combining information from the trade data and the consumer data, such that searches can be performed on the combined data without joining trade data and consumer data at query run time. The benchmarking query application executes queries on the data file generated by the build computer.