Most job-related decisions are based on information concerning the nature or structure of work. For example, the extent to which certain aspects of the job are separable and critical for successful performance often serve as inputs to an organization's performance management and compensation systems. Because the information needs to be as veridical as possible, it is common to have this information provided on multiple perspectives of the job and from multiple sources. This multimethod-multirater structuring of job data is typically analyzed to determine such things as convergent and discriminant validity. It is imperative that the user select the correct model for these analyses, as each approach makes very different assumptions about the composition of the data. The two most common approaches - the additive and direct product models - are described. Data from 7 different jobs are analyzed. It is argued that the direct product model provides a better representation of the data.