BlackRock has launched a new research framework to improve data quality in financial modeling, targeting the prevalent issue of dirty data. The initiative addresses critical aspects of data handling through a structured approach that integrates quality checks at three key stages: data ingestion, model checking, and output validation.
The first step focuses on standardizing incoming vendor feeds and identifying duplicates, aiming to rectify structural issues early. Subsequent monitoring of model behavior helps detect anomalies in outputs, such as unexpected yield gaps, to maintain data integrity. Finally, a validation process ensures that only accurate data reaches decision-makers, safeguarding against erroneous signals that could disrupt trading and reporting.
Additionally, BlackRock’s AI-based data completion module has shown significant improvements, reducing false positive rates from approximately 48% to around 10%. Benchmark tests indicate the system achieves about 90% recall and 90% precision, marking a 130% improvement over previous methods. The framework is currently operational and aims to enhance the reliability of financial data management.