How a Warning Becomes Wallpaper

The Global Signal

On January 28, 1986, the Space Shuttle Challenger broke apart 73 seconds after launch, killing all seven crew members, due to the failure of an O-ring seal in one of the solid rocket boosters in unusually cold launch-morning temperatures. The Rogers Commission, the presidential commission that investigated the disaster, found that engineers at contractor Morton Thiokol had documented O-ring erosion and blow-by on prior shuttle flights for years before the accident, and had raised concerns about launching in cold temperatures the night before the flight, concerns that were ultimately overruled in the final launch decision (Rogers Commission Report, 1986). Sociologist Diane Vaughan's subsequent study of the decision, "The Challenger Launch Decision" (1996), coined the term "normalization of deviance" to describe how each prior instance of O-ring erosion, technically a deviation from the design specification, had been reviewed, explained, and accepted as within an expanding definition of acceptable risk, so that by the night before the final launch, a genuinely dangerous pattern had been absorbed into what counted as normal.

No single review of any one flight's O-ring data was dishonest or obviously wrong. Each review, taken alone, reached a defensible conclusion. The pattern only became visible in hindsight, once someone looked at the flights together instead of one at a time.

The Hidden Signal

A risk indicator that recurs across separate incidents, each reviewed and explained individually, can accumulate into a clear pattern that no single review is positioned to see, because each review's job is to explain that one instance, not to compare it against the accumulating series. Consider a hypothetical scenario, smaller than Challenger's case but illustrative of the same mechanism: a manufacturer experiences a minor product defect rate that ticks up slightly each quarter for two years. Each quarterly review explains that quarter's uptick with a specific, plausible cause, a supplier delay, a seasonal demand spike, a training gap on one shift, and each explanation is individually reasonable. No single review compares all eight quarters together, so the underlying, worsening trend is never named as a trend; it exists only as eight separately explained blips.

Why the Visible Metric Misleads

A quarterly or per-incident review answers a narrow question well, what explains this specific instance, and that narrow framing is exactly what prevents it from answering a different, more important question: does this instance belong to a worsening pattern across time. The more revealing practice is not better individual incident reviews, which were often genuinely competent on their own terms in Challenger's case, but a separate, standing review that looks specifically for whether prior "acceptable" explanations are recurring and whether the range being called acceptable has quietly widened. The Rogers Commission found that Morton Thiokol's own engineers had, in effect, already updated their internal sense of acceptable O-ring erosion multiple times before the fatal flight, each update reasonable given the immediate data, cumulatively dangerous given the full series.

Implication for Operators

Any organization that reviews risk incidents one at a time, however competently, should assume that a genuine pattern can accumulate invisibly across those reviews, because no single review is built to compare itself against the full series. The practical shift is establishing a standing, separate review whose explicit purpose is checking whether this period's accepted explanation matches or exceeds prior periods', not simply reviewing each new incident on its own merits.