We describe a method to improve credit portfolio models based on the Merton model by adding to the underlying distributions forward-looking tails deducted through the Bayesian Networks technology. Given the forward-looking stance of the approach, its results give a better quanti ed picture of the vulnerabilities of an institution under extreme stress and at the same time satisfy the Basel II recommendations for integrating forward-looking stress scenarios in the decision making process and capital planning. We show the procedure in detail in a stylized case.