Constraints to growth for medium-sized enterprises in low-income countries remain a puzzle and a challenge, for academics and policymakers respectively. We collaborated with a lender in Egypt to examine whether some borrowers already in their portfolio could make profitable use of radically larger loans. We find null average effects on borrower profitability, but see some evidence of increased household expenditure. Using psychometric baseline data, combined with machine learning methods, we find dramatic heterogeneity: firms experiencing the highest quartile of treatment effects increase profits by 55% relative to the control group; those in the lowest quartile experience a 52% reduction in profits. We characterize the psychometric questions as identifying “go-getters” and find that while go-getters experience higher enterprise profits and productivity in normal conditions (i.e., the control group), when capital constraints are relaxed (i.e., the treatment group) they become impetuous and perform worse. The cautious non-go-getters make better use of the additional capital.