It should be possible to quantify the performance of a surface inspection system with reasonable effort. This simplifies the specifications of the system, standardises the acceptance process and helps to monitor the performance of the systems during operation. If the surface inspection system is proven to operate at a specified performance, quality statements can be made about the product from the system’s data output. Even performance statements about the production system are possible via trend analyses. In VDI/VDE/VDMA 2632 Part 3.1, various methods for testing classifying machine vision systems are already described. However, these methods are not practicable with the general conditions that are typical in flat steel production. VDI/VDE/VDMA 2632 Part 4.2 now presents a procedure for the performance evaluation of inspection systems that is practicable under the general conditions of flat steel production. The standard is intended for anyone who, as a supplier or user of surface inspection systems in flat steel production, wishes to acquire the classification performance to make a performance statement about the surface inspection system. Generalised, the performance evaluation procedure can be applied to all classification tasks where the "ground truth", i.e. a statement about the true distribution of the events among the different classes, is not possible. This situation always applies in practice when the number of events to be considered is very large and it is therefore not possible to check the classification in each individual case.