Importantly, the number of counterfactual explanations may grow exponentially with respect to the number of features, so finding all possible counterfactual explanations is generally intractable when the number of features is large. In the case of the loans discussed in this empirical example, we did an exhaustive search because the number of features was relatively small, thus Tables 1-2 show all possible counterfactual explanations for the credit denials of Loan 1 and Loan 4. In most settings, however, the algorithm would need to be restricted to a maximum number of iterations or explanations to be tractable. Nonetheless, the Counterfactual Explanations for Data-Driven Decisions Fortieth International Conference on Information Systems, Munich 2019 7 number of explanations may still be large, so additional steps to improve interpretability could be helpful, such as defining measures to rank explanations according to how “good” they are.