This paper studies the determinants of financial performance (return on assets, ROA) of 18 Argentine Microfinance Institutions (MFIs) from 2002 to 2018. We apply the random forest algorithm to predict the ROA of the Argentine MFIs, introducing two social variables to capture the depth of the outreach such as the female ratio and the average size of the loan portfolio divided by the GDP per capita. We also consider five other main explanatory variables, such as the size, efficiency, quality of loan portfolio, solvency, and productivity ratio, as well as macroeconomic variables. Although our results indicate that the quality of the loan portfolio and efficiency are the most important variables in predicting ROA, we find that social variables are also important; in particular, the female ratio, which is the third relevant predictor of ROA. In contrast, macroeconomic variables and the financial crisis turn out to be insignificant in our analysis.