There is often a divide between those whose job is to promote and put into place safety interventions (people such as policy makers or company managers) and those with the expertise and interest in evaluating how effective such interventions are (people such as researchers). I will discuss some possible reasons for this later. What I want to do in this talk is begin to bridge the gap, while recognizing the difficulties to be overcome. Indeed, I raise questions as much as I answer any. I will start by very briefly describing the context in which I work and in which I have most experience. I then consider what is known about factors related to better accident performance, and how to choose appropriate interventions. I will describe the need for evaluating interventions, and then consider the quality of evaluations. This will set the scene for further discussion of the issues involved using real life examples.