Quality data about business processes in small software companies are analysed using neural network based tools. It is shown that this technology is powerful: a) to identify "types" of businesses, b) to learn functions on overall performance dependent on specific quality parameters, and c) to identify "improvement steps", i.e. which measures are most appropriate to achieve higher performance.

In the application using sample data of 50+ business units it was found that thorough inspection of early software life cycle phases contributes most to high performance. The method has been used officially in an analysis of the Swiss software industry, and it is possible to apply it to other types of business data.