This research report, produced using internal M&E data, serves as a proof of concept for the use of advanced analytics techniques to inform the screening of EDF businesses and grant modalities. We found that machine learning models can accurately predict simple measures of business success after 12 months - an increase in profit or the achievement of the firm's job creation goal - using baseline survey data collected at the time of award. By using data-driven models that reflect the diversity of business profiles, the programme could better identify high-potential applicants that may be overlooked by more traditional screening tools, while ensuring an ethical, humane and transparent selection process. Such models can also help identify sectors, demographic groups and geographical areas facing structural barriers that require complementary forms of support.