We describe an early warning system (EWS) for credit risk management based on fuzzy expert systems. Modelling process, including knowledge elicitation and univariate analysis is discussed. Several approaches to handling model complexity, model auto tuning and validation are described.

 We found that such a hybrid model had more predictive power and is more robust than a purely statistical one. Even more important is that the methodological choice facilitated the discussion between stakeholders and increased involvement, acceptance and interdepartmental communication.