For the banking system, forecasting the risk of company failure supposes that tools for detecting company difficulties which make use of widely available computer databses are available. Such tools can be used as a decision-making aid in credit arrangements on a case-by-case basis and also to manage risk arising on all the bank’s committments. This article presents: (a) the constructin of the Banque de France industry score: i.e. the choice of data and statistical method, the validation of the tool, and the estimated probability of failure according to the score function; (b) the method of diagnosing a company’s individual risk; (c) the use of the score as a probabilistic tool for evaluating the risk arising on a portfolio of commitments.