New Technologies for Customer Rating: Integration of Knowledge-Based Systems and Human Judgement

Avoiding incorrect credit decisions and minimizing the required effort are crucial factors for the economic success of a bank. A basic task of credit granting is a correct valuation of a customer's annual account. CUBUS was designed to give general valuation and a first global impression of a company quickly. It also draws the user's attention to the most critical items of the balance sheet. In addition, detailed analysis and valuations of each annual account item is provided. CUBUS goals which have been achieved include increasing decision quality, improving the know-how level of the staff, improving communication among different departments and finally defining and establishing generally accepted guidelines and norms. The system allows very flexible human-computer cooperation. It works as a competent assistant which provides the following functions: autonomous valuations, reactive assistance, active consulting, interactive problem solving, tutoring and supervising. When SBC's (Swiss Bank Corporation) general management decided that all commercial customers had to be rated, an application had to be developed to make CUBUS judgements available to all credit specialists within the existing mainframe database environment. The integration of mainframe databases and transactions, knowledge-based systems running on workstations, and the judgement of human users raised the challenging question of what general principles have to be applied when such diverse components have to be combined in the most efficient way.