[Excerpt] Utility analysis has been described as 'a family of theories and measures designed to describe, predict, and/or explain what determines the usefulness or desirability of decision options, and to examine how information affects decisions' (Boudreau, 1991, p. 621). Utility analysis (UA) evolved to provide tools for better describing and communicating the impact of HRM and industrial psychology interventions on organizational goals. Most UA research has been based on an implicit assumption that such communication would be aimed toward managers, who control the resources necessary to implement such interventions. Previous research has noted the need to move beyond developing new measures of utility parameters, and to focus on the role of UA information in managerial decision processes (Boudreau, 1989, 1991). Yet there remains a lack of research exploring these issues. In this chapter, we attempt to explicate some of these underlying assumptions and suggest how future UA research may fruitfully test them. We focus specifically on areas where the practice of human resource management seems to diverge from the implied behaviors of UA theory. These deviations provide potential clues about how to enhance the accuracy and usefulness of UA models, and suggest new directions for future UA research and practice. The themes developed in this chapter relate to UA applications in training, compensation, performance assessment, and internal staffing. However, for simplicity and exposition, we will use the external selection model as our guiding framework. External selection is the focus of a significant proportion of UA research, so using this model will relate our themes to a large body of literature. Using the external selection model, this review will attempt to 'open up the black boxes' of four fundamental UA processes: 1. The relationship between predictors and criteria, represented by r, the correlation coefficient. 2. The nature of the criteria, represented by SDy. 3. The nature of the selection process, represented by Zx. 4. The nature of the implementation process, represented by C.