The NASA Lesson Nobody
Learned
On January 28, 1986, the Space Shuttle Challenger disintegrated 73
seconds after launch, killing all seven crew members. The technical
cause was clear: an O-ring seal in the right solid rocket booster failed
due to unusually cold temperatures. But the organizational cause was far
more insidious and far more relevant to every quality professional
reading this.
The O-ring erosion had been observed on previous flights. Flight
after flight, the seals showed signs of blow-by and erosion. Each time,
engineers noted it, discussed it, and ultimately accepted it. The
reasoning was always the same: “It eroded, but it didn’t fail. So the
erosion must be within acceptable limits.” What began as an anomaly
became an expected observation. What was once a deviation became a
parameter. What should have been a red flag became a line on a chart
that nobody flinched at.
Diane Vaughan, the sociologist who studied the Challenger disaster in
depth, gave this phenomenon its name: the normalization of
deviance. It describes the process by which individuals and
organizations gradually accept lower standards of performance, safety,
or quality until those lower standards become the new normal. The
deviation doesn’t happen overnight. It happens one small exception at a
time, each one reasonable in isolation, each one building on the last,
until the organization is operating in a state that would have been
unthinkable just months or years earlier.
And here is the uncomfortable truth for every quality manager,
manufacturing engineer, and operations director: your organization is
doing this right now. Maybe not with O-rings. But with something.
What
Normalization of Deviance Looks Like in Manufacturing
Normalization of deviance doesn’t announce itself. It doesn’t show up
as a dramatic policy change or a bold decision to lower standards. It
arrives as a series of quiet, pragmatic compromises that each make
perfect sense in the moment.
Consider a CNC machining operation. The specification calls for a
surface finish of Ra 0.8 μm. One day, a part comes off the line at Ra
0.95 μm. The operator flags it. The quality engineer investigates and
finds that the cutting tool is slightly worn. Replacing it would require
stopping the line for 45 minutes, and there’s a shipment due tomorrow.
The engineer makes a judgment call: “It’s close enough. The functional
impact is minimal. Let’s approve this one and change the tool at the
next scheduled maintenance.”
Nobody writes a deviation report. Nobody updates the control plan.
Nobody conducts a root cause analysis that traces back to tool change
intervals. The part ships.
Two weeks later, another part comes in at Ra 0.92 μm. Same situation,
different operator, different quality engineer on shift. But this time,
there’s precedent. “We accepted 0.95 last time. This is actually better
than that.” The part ships without even a conversation.
Three months later, parts are routinely coming off at Ra 1.0 to 1.1
μm. The operators have stopped flagging them. The quality engineers have
stopped asking questions. The surface finish specification on the
drawing still says Ra 0.8 μm, but the operational standard — the real
standard, the one people actually follow — has drifted by 37%. And
nobody can tell you exactly when it changed, because it changed one
reasonable decision at a time.
Now apply this pattern to every dimension, every material property,
every process parameter, every inspection criterion on your production
floor. How many “reasonable exceptions” have you accumulated? How many
of your specifications exist only on paper while a different, unwritten
standard governs what actually gets produced?
The Psychological Mechanics
Understanding why normalization of deviance happens is essential to
preventing it. The process follows a predictable psychological
pattern.
First, there’s the initial deviation. Something
doesn’t go according to plan. A parameter is out of spec, a step is
skipped, a tolerance is exceeded. This triggers discomfort, discussion,
and usually a conscious decision to accept the deviation based on
circumstances.
Second, there’s the absence of consequences. The
deviant part is installed, the skipped step doesn’t cause an immediate
problem, the out-of-spec parameter doesn’t result in a customer
complaint. This absence of negative consequences is catastrophically
important. The human brain interprets “nothing bad happened” as evidence
that the deviation was safe, rather than evidence that the organization
got lucky.
Third, there’s the reinterpretation of risk. Because
nothing bad happened the first time, the second occurrence feels less
risky. And the third even less so. The organization’s internal risk
assessment, never formally updated, gradually shifts through accumulated
anecdotal experience. What was once a “deviation” becomes a “known
condition.” What was once a “known condition” becomes a “standard
operating parameter.”
Fourth, there’s the institutionalization. New
employees are trained by people who have already normalized the
deviation. They learn the real standard — the operational one, not the
documented one — from day one. They never even know they’re operating
below the original specification, because the deviation predates their
arrival. The organization has literally forgotten what “good” used to
mean.
Fifth, there’s the defensive rationalization. By the
time the deviation has become institutionalized, anyone who questions it
faces a formidable defense: “We’ve always done it this way.” “We’ve
never had a problem.” “The spec is probably too tight anyway.” The
deviation has become identity. Challenging it is no longer a technical
discussion; it’s a criticism of the team’s competence and history.
The Warning Signs
Normalization of deviance leaves footprints. If you know what to look
for, you can catch it before the catastrophe arrives.
Informal tolerance ranges that differ from documented
specifications. Ask your operators what the “real” limits are.
If the answer differs from what’s on the drawing, you have normalized
deviance. The existence of an unwritten, oral tradition of tolerances
that are “actually OK” is the single most reliable indicator.
Patterns of waivers and concessions. If you’re
granting deviations for the same parameter, on the same product, more
than twice, you’re not managing exceptions. You’re managing a new
standard that you refuse to acknowledge. Every recurring waiver is a
specification revision that someone is too uncomfortable to make
official.
Declining rates of nonconformance reports without
corresponding process improvements. If your NCR rate drops but
you haven’t changed your process, your people have stopped reporting.
They haven’t stopped finding deviations; they’ve stopped considering
them deviant. A quality system that reports perfection is often a
quality system that has redefined perfection downward.
Experienced personnel dismissing concerns from
newcomers. When a new hire flags something and a veteran says,
“Don’t worry about that, it’s always like that,” the normalization is
complete. The new person is being initiated into the deviant
standard.
Audit findings that reference “historical
precedent.” If the justification for accepting an out-of-spec
condition is “we’ve accepted this before,” you are building your quality
case on accumulated deviance rather than engineering analysis.
The Manufacturing
Examples Are Everywhere
This isn’t theoretical. Normalization of deviance has been implicated
in some of the most significant quality failures in industrial
history.
Boeing’s 737 MAX. The Maneuvering Characteristics
Augmentation System (MCAS) was designed to activate in specific flight
conditions. Through iterative design decisions, the system’s authority
was increased, its redundancy was reduced, and its failure modes were
downplayed. Each change was rationalized. The cumulative result was two
fatal crashes and a worldwide grounding. The normalization happened not
through one dramatic decision but through a cascade of small
compromises, each defensible in isolation, each accepted because the
previous one hadn’t caused a problem — yet.
The Takata airbag recall. Over years, Takata
gradually accepted deviations in propellant manufacturing. Moisture
content, density variations, and environmental controls during
production all drifted from their original specifications. The
organization normalized these deviations because the airbags deployed
successfully in the vast majority of cases. The failures — explosive
ruptures that shot metal shrapnel into passengers — were rare enough to
be treated as anomalies rather than as the inevitable consequence of a
cumulative drift away from the original standard. Over 100 million
airbags were recalled. At least 27 people died.
Pharmaceutical manufacturing. In 2012, the fungal
meningitis outbreak traced to the New England Compounding Center killed
64 people. The investigation revealed that the facility had normalized
practices that violated basic pharmaceutical manufacturing standards:
unclean rooms, expired sterilization processes, and ignored
environmental monitoring data. Each deviation was small. The cumulative
result was contamination that killed.
How to Fight Back
Combating normalization of deviance requires deliberate, structural
interventions. Good intentions and heightened awareness are not enough;
the psychological pressures that drive normalization are too strong.
Establish and enforce red lines. Not every parameter
needs to be a hard specification, but the ones that do must be
absolutely inviolable. If a parameter is critical to safety or function,
there should be zero tolerance for deviation without a formal,
documented, engineering-reviewed process. The key word is “formal.”
Informal acceptance is where normalization begins.
Implement periodic standard recalibration. Every
quarter, pull the original specifications for your top 10 products and
compare them to what’s actually being produced. Not to what’s being
reported — to what’s actually on the parts. Use measurement data, not
inspection pass/fail records. You will be surprised how often the
operational standard has drifted from the documented one without anyone
noticing.
Rotate quality personnel. The people who normalized
the deviation are the least likely to see it. Bring in fresh eyes — from
other product lines, other facilities, or even external auditors —
specifically tasked with comparing actual practice to documented
standard. The question isn’t “Is this acceptable?” The question is “Is
this what the specification says?”
Create a deviation log that tracks patterns. Don’t
just record individual deviations. Analyze them for recurrence. If the
same type of deviation appears three or more times, treat it not as
three separate exceptions but as one specification that needs to be
either formally revised or formally enforced. The choice between those
two options is an engineering decision. The failure to make the choice
is a management failure.
Reward the people who report. If your culture
punishes or even subtly discourages people who flag deviations, you are
actively breeding normalization. The operator who stops the line to
report an out-of-spec condition should be thanked publicly. The engineer
who refuses to sign off on a questionable part should be supported, not
overruled by someone with a shipping deadline. If the people who uphold
the standard are treated as obstacles, the standard will not
survive.
Distinguish between innovation and deviance. Not
every deviation from a specification is bad. Sometimes processes improve
and specifications need updating. But there is a profound difference
between a deliberate, documented, analyzed change to a process and an
informal, undocumented, unanalyzed drift. The former is continuous
improvement. The latter is normalization of deviance. The difference is
rigor, documentation, and intent.
The Uncomfortable Mirror
If you’re being honest with yourself, you can probably identify at
least one area in your organization where the operational standard has
drifted from the documented one. Maybe it’s a cleaning procedure that
gets shortened on the night shift. Maybe it’s a calibration interval
that gets extended because the instrument “always passes anyway.” Maybe
it’s an inspection step that got dropped during a staffing shortage and
never reinstated.
Each of these seems small. Each feels manageable. Each has a
perfectly reasonable explanation. And each is a step on the path that
NASA walked, that Boeing walked, that Takata walked.
The normalization of deviance doesn’t require bad people making bad
decisions. It requires good people making reasonable decisions in the
absence of structural safeguards. It requires intelligent professionals
who interpret the absence of failure as evidence of safety rather than
evidence of luck.
Your quality system is either drifting or it isn’t. If you think it
isn’t, look harder. If you find that it is, the time to act was before
the deviation became normal. The second-best time is now.
The catastrophe you prevent won’t make the news. There will be no
congressional investigation, no viral media coverage, no dramatic
hearings. The absence of disaster is, by its nature, invisible. But that
invisibility is exactly what makes the prevention so critical — and so
easy to neglect.
Every deviation you accept today is a standard you’ll defend
tomorrow. Choose carefully.
Peter Stasko is a Quality Architect with over 25
years of experience in manufacturing excellence, process optimization,
and quality management systems. He has helped organizations across
automotive, aerospace, and industrial manufacturing build quality
cultures that don’t just detect defects — they prevent the conditions
that allow defects to exist.
