A small Grouchy Smurf figurine with a butterfly on its nose sits on Jeff Caldwell’s desk, and not necessarily because he’s a fan of the blue forest-dwelling sprites.
Rather, Caldwell keeps it there because it was a gift from one of his teams as a reminder of a quality turnaround they all worked together on earlier in his career. During a visit to a manufacturing facility regulated by the US Food and Drug Administration (FDA), he immediately noticed it was dirty and overflowing with nonconforming material, and he later learned the company’s customers were unhappy.
Yet for some reason, the people who worked there only saw blue skies for miles.
“The employees were all like happy little Smurfs,” recalls Caldwell, a QualityHub Principal Consultant. “Nobody experienced any pressure. Nobody experienced any problems. In their eyes, everything was good. And nothing ever impacted them negatively – or so they thought.”
Caldwell said that’s where the figurine comes in.
“Every village needs a Grouchy Smurf – the one who looks around while everyone else is singing and says, ‘I don’t like this,’” he said. “In a quality system, somebody has to be willing to be that person and point out what everyone else has stopped seeing. The butterfly is my reminder that you can do it without being miserable about it.”
“People always talk about biases as a bad thing. It’s not necessarily a bad thing. The trouble starts when people apply their biases carte blanche and never challenge them.” – Jeff Caldwell
That blind spot – the gap between what those workers saw and what Caldwell saw – has a name: “normalization of deviance,” a term introduced by sociologist Diane Vaughan in her study of NASA’s fatal 1986 Challenger launch decision. Vaughan described how an organization can come to accept a risky or improper practice as routine simply because repeated departures from the standard haven’t yet ended in disaster.
The shift happens slowly. A team skips a step, bends a requirement, or tolerates an anomaly, and nothing goes wrong. Each uneventful outcome makes the next exception easier to justify, and over time the lowered standard quietly becomes the accepted way of working. What once would’ve raised concern no longer registers as a problem at all.
Caldwell said normalization of deviance is one of the least discussed and most consequential risks in regulated manufacturing. In medical device and pharmaceutical companies, it can quietly move a business from minor nonconformances to FDA Form 483 inspectional observations, warning letters and, in the worst cases, a consent decree, along with costly product recalls in some cases.
A Bias, Not a Character Flaw
Caldwell frames the condition in psychological terms, a view shaped by years of turnaround work and a neuroscience certificate course he completed along the way. People form biases from their experiences and opinions, and those biases shape the decisions they make. In a regulated environment, that means gradually accepting conditions of noncompliance because nothing bad has happened – yet.
“We normalize ourselves to conditions where there’s noncompliance,” Caldwell told QHub Insights. “You may not have a negative impact immediately, but it doesn’t remove the likelihood that it’ll happen eventually, and you progressively accept worse and worse risk until the negative impact occurs.”
He said bias itself isn’t the enemy. Biases help people make decisions faster than they could without that experience.
“People always talk about biases as a bad thing. It’s not necessarily a bad thing,” Caldwell said. “The trouble starts when people apply their biases carte blanche and never challenge them.”
For quality professionals, he said, it’s a question of decision-making effectiveness and meeting requirements. Even the most conscientious person has biases and more to learn.
The Near-Miss Trap
Of all the biases he has seen affect quality systems, Caldwell points to near-miss bias as the most common. The reasoning runs along these lines: I know there’s a problem, but I’ve never felt its consequences, so I don’t expect to feel them next time.
He sees it play out most often around audits. “An auditor comes in and doesn’t catch the nonconformances. The report shows no findings, and senior leadership concludes there’s no problem.”
Each clean report feels like proof of safety, so when a new employee later flags a long list of nonconformances, leadership pushes back and points to years of audits by a certification body that never found them.
Then the FDA arrives with a different perspective.
Caldwell said the difference is in what each assessment is designed to do. A certification audit evaluates conformity to a standard and results in a certificate. Meanwhile, an FDA investigator is evaluating whether a company is in a state of control under US law.
“It’s much more effective for somebody to see and experience the issue than to just talk about it.” – Jeff Caldwell
For medical device manufacturers, that means the Quality Management System Regulation (QMSR) at 21 CFR Part 820, which incorporates international quality systems standard ISO 13485:2016 by reference, while for makers of drugs, it means the current good manufacturing practice (CGMP) regulations at 21 CFR Parts 210 and 211.
A shared standard, such as with MedTech under the QMSR and ISO 13485, doesn’t make assessments of the two the same. They differ in purpose, authority, and consequences, and a clean certificate isn’t evidence of a state of control.
“The leader thinks, ‘Well, I have an audit program in place, so I don’t have a problem,’” he said. “But internal quality audits and regulatory audits or inspections are measuring two different things, two different ways, which can cause leaders to interpret them as being the same.”
Incentives can reinforce the drift without anyone intending it. When audits come back clean, metrics look healthy, and bonuses keep arriving, there’s little built-in reason to look for problems the audit didn’t surface. Over time, the absence of bad news can start to feel like confirmation that there’s nothing to find.
That’s why Caldwell sees serious enforcement actions as the end of a long chain of ignored regulatory signals. He says a consent decree happens “because people haven’t reacted to the signals, because they haven’t felt the negative impact. …Yet.”
What an Investigator Notices First
Caldwell said good FDA inspection preparation comes down to showing that a business is in a state of control. Many small signals can lead an investigator to decide early that a business is not in control, and that management doesn’t recognize it.
The earliest warning sign isn’t the problem itself, he said. It’s when the people who work there don’t notice it. “I see a dirty vent. But you don’t see a dirty vent,” Caldwell said. “You don’t even recognize the situation, and that’s where [an investigator] starts forming their opinion.”
The gap shows up in conversation too. An experienced auditor or investigator is looking for confirmation of control. If the people answering questions explain why a noncompliance is acceptable instead of how they’ll fix it or acknowledging that it was worth noticing, that tells the investigator something.
“An investigator is there to evaluate whether you’re achieving control, and if you’re explaining why it’s OK to have a problem, then your biases are clearly different.”
That’s why he sees the quality professional’s role as part watchdog and part translator. In his words, the job is to “look for areas where we may be broadcasting an impression that we are not in control as a business,” and then help leadership and staff understand when their personal biases are working against a state of control.
Those signals often show up in a company’s own data before an investigator ever arrives, including:
- Deviations that keep recurring;
- A growing share of nonconforming material dispositioned “use as-is” or released under concession;
- Nonconformances and corrective and preventive actions (CAPAs) that linger for months without meaningful progress;
- CAPAs closed without any check that the fix worked; and
- An audit program that reports zero findings year after year.
None of them alone proves a problem, but together they suggest a quality system has begun accepting what it once would’ve challenged.
Seeing is Believing
Asked what he would say to a plant manager who is sure their site doesn’t have a problem, Caldwell answers frankly.
“There’s not a lot you can say,” he said. “It’s much more effective for somebody to see and experience the issue than to just talk about it.”
Caldwell suggests a Gemba walk of the facility with management, pointing out areas of noncompliance and tying what is found to the state of compliance. Repeat it with a team, he said, and “everyone frames to that new bias, and they see the disconnect.”
In that turnaround he described with the “happy Smurfs,” photographs were Caldwell’s most effective tool. He walked the facility for an hour, took pictures, and presented them at an all-hands meeting under the title “This is What the Customer Sees.”
The photos kept the conversation from becoming personal.
“I didn’t allow it to be a case of ‘me against them,’” Caldwell said, noting that his message to employees is that the condition of the facility reflects the systems and expectations of the people who work within. “I took photographs of the situation and I put it in front of them and asked, ‘Do you see what I’m seeing? Is that acceptable to you?’”
Management Review as a Check on Bias
A Gemba walk can reset people’s perspectives, but QualityHub’s Caldwell warns that the reset doesn’t always last. Workers tend to regress to their earlier view because that’s what’s been ingrained in their minds.
Caldwell said bias is best kept in check by bringing people together, because leaders with different perspectives tend to challenge one another’s blind spots. For that reason, he called management review one of the most useful and least appreciated tools a quality system has. Each functional leader brings biases formed by their own experience, and a well-run management review aligns the team on how the quality system is interpreted and carried out. (Related Story: “Successful MedTechs Use Management Controls to Oversee Quality Systems, Avoid FDA Risk. Here’s How Yours Can, Too,” QualityHub, Dec. 4, 2025.)
But that alignment has to work in both directions. Caldwell said quality professionals should recognize that their own bias for compliance isn’t the only valid perspective in the room. Left unchecked, he noted, the quality bias would be to avoid shipping nonconforming product by never releasing anything, and the business would fail. In a group setting, the quality leader may be recalibrated toward a position the organization can support that still meets requirements.
Caldwell added that it matters how the message is framed. “If you’re talking to a finance person, don’t talk about CTQs,” or Critical to Quality characteristics, he said. “People will get there eventually, but quality leaders are more effective when they show how compliance supports the goals each function already cares about.”
To keep small deviations from becoming culture, he points to truly independent audits, not audits by people who are part of the same environment and share its biases.
An audit that is graded on a curve serves no one because the goal is to find out what’s really happening, Caldwell said. Once the results are in, the quality professional should help leadership understand how they differ from the company’s assumptions, and do it in a way that points to clear action.
Change Happens Over Time
Caldwell said quality leaders should expect resistance, especially in turnarounds.
After all, “the Smurf Village was always happy before Gargamel showed up,” he said, stressing that behavior doesn’t change right away.
At the company where Caldwell began with photographs, the work took a few years. By the end, the company’s performance had improved enough to seize market share from a competitor based on a strategy aligned with the highest quality expectations. And on his last day there in a professional capacity, some of the employees who initially pushed back hardest thanked Caldwell and shared with him that they believed they still had jobs because of what he taught them. At that point, they also gifted him with his Grouchy Smurf figurine to remember their experience together.
For Caldwell, that experience reinforced one lesson: people who can’t see a problem are rarely bad people. Neither are the “new people” who join the team and may see different risks and perspectives.
“It’s because we have our own biases,” he said. “Left unchallenged, those biases may normalize the problem, and eventually it reaches the patient.”
