Nowadays, complex electronic products, such as DVD players or mobile phones, offer a huge number of functions. As a consequence of the complexity of the devices, customers often have problems to use such products effectively. For example, it has been observed that an increasing number of technically sound products is returned due to, e.g., interaction problems. One possible root cause of this problem is that most product development processes are still too technologydriven, i.e., potential users are brought into contact with the product only at a very late stage. If early consumer tests are carried out, then these typically aim at abstract market evaluations rather than formulating concrete requirements towards the functionality of the product. As a result, products often have little meaning or relevance to the customers. Therefore, we need better ways to involve users in the development of such products. This can be achieved by observing product
usage in the field and incorporating the gained knowledge in the product creation process. This paper proposes an approach to build automatic observation modules into products, collect usage data, and analyze these data by means of process mining techniques exploiting a novel semantic link between observation and analysis. This link yields two main benefits: (i) it adds focus to the potential mass of captured data items; and (ii) it reduces the need for extensive post-processing of the collected data. Together, these two benefits speed up the information feedback cycle towards development.