page numbers refer to Internet reproduction of actual article

p.1 "Slack." The term reeks of inefficiency. It implies a looseness in the organization of things--one part of a system trailing badly the activity of another. Yet slack is a critical, if underappreciated, managerial resource. It is slack that provides a margin for error between a lapse in one aspect of an organization's performance and harmful consequences in every other. It is slack that allows managers the freedom to maneuver--to act decisively on one part of a problem without having their decisions ramify quickly, widely, and unexpectedly to every other part.

p.1 Reliability is a complex and difficult organizational trait to achieve. In pursuit of it, many managerial instincts push precisely toward strategies that threaten even the last vestiges of organizational slack.

p.6 At Diablo Canyon, there is a widespread recognition that all of the potential failure modes into which highly complex technological systems could resolve themselves have yet to be experienced. Nor have they been exhaustively deduced. In this respect, the technology is still capable of surprises.

p.8 Consider briefly two theories of reliability that could be applied to the performance of complex, hazardous operations (Wildavsky, 1989). Under one theory, reliability would stem from a constant, certain, predictable set of performances. All system conditions would be fully specified and anticipated. Term this the "anticipatory model" of reliability--an approach that equates reliability to invariance. Here an organization ought to determine its functions, or at least strive to determine them, unambiguously and completely. Once "correct" job performance is specified, it should be "locked in" once and for all through formal procedures, unvaryingly applied. A unified chain of command guarantees swift action and preserves the "perfect" model.

But a second approach to reliability is possible. Here reliability would be equated not with invariance but with resilience. Being responsive to, rather than trying to weed out, the unexpected would be the ultimate safeguard of stable performance. Despite the knowledge and elaborate procedures of an organization, its technology, it would be believed, is still capable of surprises. This expectation of surprise would not only be a state of mind, it would be recognized as an important organizational resource. Under this model of reliability, an organization would value its capacity for real-time discovery as much as its ability to control by anticipation.

p.9 The Diablo Canyon case is an important example to other organizations that, under diminished slack, attempt to lock in "safe" levels of performance.

p.10 The lesson from Diablo Canyon is that a core capability of adaptation and resilience is the key to maintaining organizational reliability at the highest levels.