ISTARI enabled OECD to predict high-growth firms using machine learning trained on historical web data. By enriching datasets with archived website content and time-based patterns, we delivered a reliable and scalable model for dynamic economic forecasting.
Limited ability to accurately forecast high-growth firms
The OECD wanted to better understand which companies were likely to exhibit significant growth, but faced major obstacles. Manual methods lacked scalability and insight, and there was limited access to suitable historical data for training predictive models. Without a data-rich, time-based foundation, building a reliable system for forecasting high-growth firms remained out of reach.
Training machine learning models with archived web data and metadata
ISTARI began by enriching the OECD’s company list with archived web data collected through intelligent web scraping. We extracted not only content but also metadata and timestamps, which enabled time-series analysis. Using this enriched dataset, ISTARI developed multiple machine learning models to predict high-growth indicators, then tested and refined them to identify the most accurate approach based on real historical patterns.
Scalable growth prediction powered by historical insights
OECD now has a robust and scalable machine learning model built on rich historical web data, enabling smarter economic forecasting. The final model was delivered to the client for in-house use. It allows OECD teams to run predictive analyses on their company data with greater accuracy and confidence. By leveraging web data and AI, they have dramatically reduced manual effort while gaining deeper insight into firm-level growth potential.
