Nobody Ever Gets Credit for Fixing Problems That Never Happened: Creating and Sustaining Process Improvement
Repenning and Sterman's account of why so few process improvement programmes produce significant results, even though the techniques often have useful content. Drawing on what they describe as over a dozen in-depth case studies across a decade, they argue the failure has little to do with the tool chosen and is rooted in how a programme interacts with the structures around it, and model it as a system dynamics causal loop diagram. Managers facing a performance gap can push people to work harder, which raises output at once, or to work smarter, which builds capability only after a delay and with some risk. Under pressure, people cut back improvement and maintenance; because capability erodes slowly, these shortcuts look effective for a while, a better-before-worse pattern that draws organisations into what they call the capability trap.
The trap persists through what they call the self-confirming attribution error. Managers prone to the fundamental attribution error, blaming people rather than the system, turn up the pressure; throughput rises, partly because workers quietly cut improvement time, so the pressure appears to work and the suspicion of slack effort is confirmed. When the hidden shortcuts come to light, managers' view of workers as untrustworthy is confirmed and monitoring tightens. In every organisation they studied, engineers routinely neglected to reveal serious design issues for fear of retribution: one firm's engineers had the motto never reveal you have a problem until you also have the solution, and another's called the weekly progress review the liars' club. The title is an auto engineer's remark, quoted where they argue that most organisations reward last-minute heroics over the work that prevents crises.
The bridge to psychological safety is ours: without using the term, the paper shows silence about problems being produced by pressure and misattribution, then read as confirmation that staff cannot be trusted. Its limits: the model is presented as general but first emerged from two initiatives at one automaker, the simulations use a hypothetical process, and the evidence is illustrative fieldwork plus two success cases, Du Pont and BP's Lima refinery, the Lima figures sourced to conference talks and personal communication.
Save this paper to a bibliography, keep notes on it, and export citations. Create a free account →