Given limited resources the main problem in designing tools for computer-supported work environments is to quickly find both a level of representation and a level of analysis that effectively explains the uncertainty that exists within the resulting "electronic space". This is also known as "the model class discovery dilemma": the problem of finding a fit between design and environmental variability. The traditional design paradigm for the field of computersupported cooperative work (CSCW) has proposed that modelers use either critical incidents or some sort of design rationale to guide the users of these systems toward productive synchronous and asynchronous work behaviors. A critique of the critical incident and design schema methods is presented based on a case study of my designing a distributed CSCW system in the customer service department of a large health maintenance organization. A third alternative that looks at designing based on more fundamental "metrics of coordination" is also presented. The implications that computational systems utilizing new coordination metrics might have on the sociotechnical work structure of the future is also presented.