A medical device manufacturer experienced a catastrophic autoclave failure that sent unsterilized surgical instruments into operating rooms across three countries. The fallout included regulatory investigations, product recalls, and a CEO resignation. Within weeks, the company spent over €4 million on two new autoclaves with redundant temperature monitoring, automated lockouts, and real-time alerting. Every related SOP was rewritten and operators were trained on sterilization protocols four times in a single year.
Meanwhile, 300 metres down the hall, the packaging line had been running with a misaligned seal bar for eleven months. The defect rate on seal integrity was 3.2%, meaning roughly 1 in every 31 packages had a compromised sterile barrier. Nobody noticed. The packaging line's control charts hadn't been reviewed in two quarters, and the last audit of the sealing process was fourteen months overdue. The autoclave failure was vivid, emotional, and unforgettable. The packaging defect was invisible, statistical, and boring.
This is the availability heuristic at work: the tendency to judge the likelihood or importance of something based on how easily examples come to mind. Identified by Amos Tversky and Daniel Kahneman in 1973, this mental shortcut compounds across teams, departments, and leadership levels until it reshapes entire quality strategies. Organizations systematically over-invest in preventing the failures they remember and under-invest in preventing the failures they don't.
How Memory Distortion Reshapes Quality Priorities
The most common form is recency distortion. A failure that happened last week looms larger than one that happened last year, even if the older failure was more severe, more costly, or more likely to recur. Quality teams pour resources into preventing a repeat of the most recent headline while ignoring dormant risks that are quietly building toward the next crisis.
I audited an automotive supplier that had received a customer complaint about surface finish defects on a visible exterior trim component. The defect was purely cosmetic, but the customer's purchasing director personally rejected the shipment on the loading dock. The story became legendary inside the plant. For the next eighteen months, every improvement project and capital request connected to surface finish. The plant invested in new polishing equipment, hired a specialist, and implemented 200% inspection on every visible surface.
During that same period, the plant's dimensional capability on a critical mounting hole drifted from a Cpk of 1.67 to 0.89. The data was sitting in the SPC system with an unmistakable trend, but dimensional capability wasn't the story anyone told around the coffee machine. When the customer eventually initiated a formal audit triggered by field failures all related to the mounting hole dimension, the quality director was genuinely shocked. They had been so focused on surface finish that the structural failure became invisible.
Dramatic failures are more memorable than gradual ones. A machine crash that stops production for three days commands more attention than a slow process drift producing thousands of marginally non-conforming parts over six months. The crash is a story with characters and conflict. The drift is a data point in a spreadsheet nobody opens.
The Structural Consequences for Quality Systems
When the availability heuristic goes unchecked, it shapes organizational architecture. Internal audit programs become backward-looking, focusing on areas where problems were found last time rather than where problems are most likely to emerge next. The audit plan becomes a tour of the organization's scar tissue instead of a systematic assessment of its risk landscape.
Training programs become reactive. Operators receive extensive training on the most recent failure topic while foundational skills, the ones that prevent slow and invisible failures, get deprioritized year after year. Capital investment follows the same pattern: equipment budgets prevent the dramatic failure everyone remembers while unglamorous infrastructure upgrades that would prevent quiet cumulative failures get deferred indefinitely.
Risk assessments become popularity contests. In PFMEA sessions, the failure modes that generate the most discussion get assigned the highest RPN scores, regardless of actual probability or severity data. The failure modes that are statistically more significant but emotionally forgettable get under-scored and under-addressed. KPI selection follows the same logic, tracking what has gone wrong rather than what could go wrong.
The cumulative effect is a quality system optimized for the past rather than the future. Resources cluster around remembered failures while statistically significant risks go unmonitored, unfunded, and unaddressed until they become the next crisis that everyone remembers.
Memory-Driven vs Data-Driven Quality Decisions
What teams do under availability bias
- Audit areas where failures were found last year, not where risk is highest
- Assign RPN scores based on discussion intensity, not occurrence data
- Fund capital projects preventing the most memorable failure
- Train operators on the latest headline defect
What a data-driven system does
- Audit high-risk and long-untouched processes regardless of recent incidents
- Rate severity and occurrence against SPC data and capability indices
- Allocate budget by comparing expected risk reduction across alternatives
- Maintain foundational skill training alongside targeted corrective training
The Aerospace Supplier: A Case Study in Misallocated Resources
An aerospace machining supplier experienced a memorable incident where the wrong material was used for a critical structural component. The part made it through several processing steps before being caught at final inspection. The cost was significant, but the real impact was reputational: the prime contractor issued a formal corrective action request under AS9100 requirements and downgraded the supplier's quality rating.
The supplier responded aggressively. They implemented material verification at receiving, at storage, at kitting, at machine setup, and at first article inspection. Five separate verification points for a failure that had happened once in the company's twenty-five-year history. The system consumed more resources than the entire tool management improvement initiative.
Meanwhile, their tool management system was essentially manual. Tool life tracking depended on operators recording usage in a logbook. Tool wear-related dimensional deviations were the single largest contributor to scrap and rework, accounting for 43% of all non-conformances in the previous year. But tool wear failures were routine, unemotional, and unmemorable. Nobody told stories about them. They were background noise.
When I presented the data comparing resource allocation against actual risk, the plant manager studied it for a long time. He acknowledged the numbers were right but said he couldn't stop thinking about the material mix-up. It was embarrassing. That is the availability heuristic in a single sentence: a near-zero probability event consuming more resources than the defect source responsible for nearly half of all scrap.
Building Systems That Override Memory
The single most effective countermeasure is building systematic processes that force decisions to be driven by data rather than recall. Risk registers must be populated analytically, not anecdotally. Every potential failure mode should be evaluated against actual occurrence data, Cpk values, customer complaint trends, and process performance metrics, not against how memorable the last occurrence was.
Structured PFMEA facilitation matters here. The facilitator should explicitly address availability bias by presenting statistical data before opening the floor to discussion. Rate severity, occurrence, and detection against documented evidence. Audit planning should follow the same principle: the schedule must be driven by systematic risk assessment, not by a map of last year's findings. Include areas that haven't had problems precisely because they haven't been scrutinized recently.
Individual memory is vulnerable to availability bias. Organizational memory systems are less so, if they are designed correctly. Maintain a comprehensive failure database that records every significant quality event, not just the dramatic ones. Include slow drifts, capability degradations, near-misses, and trends that were corrected before they became crises. Make this database the starting point for every risk assessment, audit plan, and improvement prioritization.
Trend long-term data and present it alongside incident reports. When the team can see that packaging seal integrity has been degrading for six months, even without a single dramatic event, it becomes harder for the latest autoclave failure to monopolize attention. Rotating quality personnel across areas also breaks the cycle of institutional memory bias, exposing the organization to fresh perspectives and unfamiliar data.
The absence of memorable failures doesn't indicate the absence of risk. It often indicates the absence of scrutiny.
Data-Driven Risk Prioritisation Process
- 01Pull the dataGather SPC trends, Cpk degradation, scrap codes, complaint logs, and near-miss records from the failure database.
- 02Score against evidenceRate PFMEA severity and occurrence using process data, not the vividness of the last remembered incident.
- 03Compare alternativesFor any proposed investment over a threshold, document expected risk reduction against the top alternative uses of the same resources.
- 04Audit the untouchedPrioritise processes that haven't had a failure or an audit in the longest period.
- 05Run the premortemImagine a quality crisis one year out and work backward to identify causes the current risk register has missed.
Actively Seeking Invisible Risks
The availability heuristic makes you prepare for what you can see. Counter it by systematically looking for what you can't. Conduct premortems: imagine that a quality crisis has occurred one year from now and work backward to identify what caused it. This forces the team to think beyond current memory and consider scenarios that haven't happened yet.
Use negative brainstorming. Ask your team what they are not worrying about. Ask what a clever competitor would identify as your biggest quality vulnerability. These questions deliberately bypass the availability heuristic by searching for risks that aren't top of mind. Review untouched processes: identify the ones that haven't had a problem in the longest time and audit them first.
Quantify emotional decisions. When you feel the urge to invest heavily in preventing a specific failure, ask whether you are responding to data or to feeling. Emotional responses sometimes point to real risks that data hasn't captured yet, but you should pause and verify. Run the numbers. Compare the proposed investment against the statistical risk. Ask whether the same resources, deployed elsewhere, would prevent more total harm.
I worked with a quality director who developed a simple discipline. Every time someone proposed a corrective action investment exceeding €10,000, she required a one-page comparison showing the expected risk reduction from the proposed action alongside the expected risk reduction from the top three alternative uses of the same resources. She didn't always choose the alternative, but the comparison ensured that availability-driven decisions were at least informed by data.
The Gap Between Memory and Risk
The availability heuristic is a feature of human cognition, not a flaw. It evolved for environments where the most memorable events were often the most important ones. But modern quality management doesn't operate in those environments. The risks that matter most, slow process drifts, accumulating capability degradation, latent design weaknesses, are precisely the risks that don't create vivid memories.
These risks are quiet, gradual, and forgettable. The availability heuristic is systematically biased against the very failures that cause the most damage over time. Organizations that manage quality effectively aren't the ones with the best memories. They're the ones with the best systems for overcoming the limitations of memory.
Your organization's quality strategy should reflect its actual risk landscape, not its most traumatic memories. The gap between what you remember and what is really there is where your next quality crisis is quietly growing. Whether you find it before it finds you depends on whether you can see past the stories you tell yourself about the failures you remember, and start looking for the ones you've been forgetting to notice.
Trust your data more than your feelings. Look where the light isn't. Prepare for the failures nobody remembers as diligently as the ones nobody can forget. Build the systems that make quality decisions driven by evidence, not by emotion. The autoclave and the seal bar are both still running. Only one of them is being watched.