Business failure prediction has developed to an important research domain in business economics over the last decades. A tremendous number of papers drew up business failure prediction models, based on different statistical modeling techniques. The most popular once are the discriminant analysis, the logistic regression analysis as well as the alternative method of artificial neural networks. Except for the occasional one, all prediction models aim at the ultimate business failure. As to date, differentiated business failure prediction models are still lacking. Various academic researchers developed models with several stages concerning the degree of danger of going concern. An intersection between these stages and the business failure prediction models would improve and specify the early recognition of the failure. The paper presents four models that developed stages of business failure and gives an overview of the empirical methods of business failure prediction and their performance. Only a few researches performed a differentiated analysis, which are covered in the paper. Based on the empirical study of Leker (1993), the paper exploratory analyses a business failure prediction model via a discriminant analysis that integrates four stages of corporate failure. This paper shows the possible application and the potential, but poses also the problems related to the business failure prediction models, especially in a differentiated case.