99-022

ON THE FEASIBILITY OF MONITORING COMPLEX WORK: THE CASE OF SOFTWARE METRICS

Robert D. Austin

Designers of management systems often assume that the performance of individuals or groups can be measured well enough to support incentives that contribute to organizational effectiveness and efficiency. Whether this assumption holds generally has often been disputed, however. Differences of opinion typically center on whether it is generally feasible to create measurement systems that do not induce dysfunctional behavior in those being measured. Some have argued that comprehensive measurement--which precludes dysfunction by measuring all dimensions of performance that are critical to value creation--is generally feasible but that realization depends on the effective design and implementation of measurement technology. Others maintain that comprehensive measurement is not generally feasible, and that infeasibility can be inherent in organizational situations in ways that cannot be overcome by measurement technology. This paper examines and compares these two positions in the context of work that is complex and intangible (e.g., "knowledge work"). An agency model is proposed to specify the organizational conditions under which dysfunction will arise. Interviews with eight practical experts on measurement in software development are analyzed to determine whether their beliefs about measurement and dysfunction are consistent with the conclusions of the model. The analysis and interviews cast doubt on the general feasibility of comprehensive measurement in software development, and, by extension, other complex work.

TOM
43 pages

| Back to 1998-1999 Working Papers | Copyright © President and Fellows of Harvard College