Product development is recognized as a critical activity in most companies. One important aspect of product development is how to handle different kinds of variation. There are many sources of variation, even at the very beginning of the product development process, there is often variation in what needs are important to different customers. But also, there is variation in the environment where the product will be used, that should be considered during product development.

The thesis presents a framework that shows how Design of Experiments (DoE), Quality Function Deployment (QFD) and Design Reviews (DRs) can be used in product development to manage variation.

Prioritization of customer needs is one of the more important purposes of QFD. Principal Component Regression is suggested as an alternative to the standard procedure for prioritization of customer needs in QFD. This approach gives different results than the standard procedure, hut further investigation is needed.

DRs are introduced as being an important part of product development. The DR meeting is a forum to discuss questions about variation and how to develop a robust product.

In DoE most procedures are specialized for the case when there is only one response variable. A methodology, including active expert involvement, which handles the case when there are many response variables, is presented. The methodology is illustrated utilizing an actuator system for the aircraft industry that is modeled in a simulation package. The strengths of the methodology are that it involves experts early in the experimental process and that it can handle situations such as when the experts change their opinion during the process.

Separating active factors from inert ones represents an important problem in the analysis of designed experiments. A modification of one of the best procedures available for this purpose is presented. This modification allows the introduction of more elaborate and less generic domain knowledge than in the original procedure. The performance of the modified procedure is evaluated in a simulation study and it is found that considerable improvements can be made unless the introduced domain knowledge strongly contradicts reality.