Learning From Mistakes at Work: The $600,000 Lesson and What the Research Says

Learning from mistakes at work: the $600,000 lesson and what the research says

There is a story that has circulated in management circles for decades. A young IBM employee makes a decision that costs the company $600,000 (in some versions a million dollars). He is called into the office of Thomas J. Watson Sr., IBM's founder, expecting to be fired. Watson tells him he is not going anywhere: the company has just spent $600,000 on his education, and it has no intention of letting a competitor benefit from it.

The story is worth telling, and it is also worth being honest about. Peter Greulich, who has written extensively on Watson's leadership, describes it as handed down by tradition, with the who, what and where changing with every retelling while the theme never does (Greulich, Tom Watson Sr. Essays on Leadership). There is no contemporaneous document. What survives is the principle, "we forgive thoughtful mistakes", and later research on psychological safety, learning behaviour and turnover costs supports much of it. This page covers what that research says, when the principle applies and when it does not, and how to build a team where mistakes turn into training instead of turnover.

Key numbers at a glance

• Replacing an employee costs one-half to two times their annual salary, and Gallup calls that a conservative estimate (Gallup)
• Just three in ten U.S. workers strongly agree that their opinions count at work; Gallup estimates that moving that to six in ten could be associated with 27% lower turnover, 40% fewer safety incidents and 12% higher productivity (Gallup)
• Workers with higher psychological safety were less than half as likely to feel stressed in a typical workday (27% vs 61%) and a third as likely to plan to leave within a year (19% vs 57%) (APA Work in America 2024, Harris Poll, 2,027 employed adults)
• In a study of 51 manufacturing work teams, psychological safety predicted learning behaviour and performance; team confidence on its own did not (Edmondson, Administrative Science Quarterly, 1999)
• Across 180 Google teams, psychological safety was the most important of five dynamics separating effective teams from the rest (Google re:Work, Project Aristotle)

Why firing for a mistake is usually the expensive option

Watson's arithmetic holds up. Gallup puts the cost of replacing an individual employee at one-half to two times their annual salary, a figure it describes as conservative, and works through the example of a 100-person organisation with an average salary of $50,000 facing turnover and replacement costs of $660,000 to $2.6 million a year (Gallup). The dismissed employee takes the lesson with them, the replacement arrives without it, and the rest of the team learns something too: that the safe move is to hide the next mistake.

That last effect is the expensive one, and it is the part most cost calculations miss. A team that hides mistakes does not make fewer of them. It reports them later, when they are larger, and it stops volunteering the near-misses that would have prevented the next one.

What the research says: psychological safety

The academic term for a team where mistakes can be admitted is psychological safety. Amy Edmondson introduced the construct in 1999 as "a shared belief held by members of a team that the team is safe for interpersonal risk taking". In her study of 51 work teams at a manufacturing company, psychological safety predicted learning behaviour (asking for help, admitting errors, seeking feedback, experimenting) and through it team performance, while team efficacy, the group's confidence in its own ability, did not once psychological safety was controlled for (Edmondson, 1999, Administrative Science Quarterly).

Google tested the same idea at scale. Project Aristotle studied 180 teams (115 engineering project teams and 65 sales pods) across a mix of high and low performers, and found that who was on the team mattered far less than how the team worked together. Of the five dynamics it identified, psychological safety, whether members feel safe taking an interpersonal risk, came first in order of importance, ahead of dependability, structure and clarity, meaning and impact (Google re:Work).

The population-level data agrees. The American Psychological Association's Work in America 2024 survey, run by The Harris Poll among 2,027 employed U.S. adults in March and April 2024, split workers at the median on psychological safety. Those above it were far less likely to feel tense or stressed in a typical workday (27% versus 61%), far less likely to plan a job search within the year (19% versus 57%) and far more likely to say their employer offers opportunities to receive feedback (57% versus 37%). The survey explicitly lists "if an employee makes a mistake, it tends to be held against them" among the markers of a psychologically unsafe workplace, and 15% of workers described theirs as toxic (APA).

Gallup's estimate of the upside is the most quoted: only three in ten U.S. workers strongly agree their opinions count at work, and Gallup estimates that raising that to six in ten could be associated with 27% lower turnover, 40% fewer safety incidents and 12% higher productivity (Gallup).

Not every mistake deserves forgiveness: the blameworthy spectrum

Watson's phrase was "thoughtful mistakes", and the qualifier matters. In Harvard Business Review, Edmondson lays out a spectrum of reasons for failure, from deliberate deviance and inattention at one end, through lack of ability, inadequate process and task complexity, to uncertainty, hypothesis testing and exploratory experiments at the other. The closer a cause sits to the deviance end, the more blameworthy it is; the closer to the experimentation end, the more praiseworthy. Her observation from working with executives is that they judge a small minority of failures to be truly blameworthy but treat the large majority as if they were, which is why people stop reporting them (Edmondson, Strategies for Learning from Failure, HBR, 2011).

The useful question after a mistake is therefore not "who did this?" but "which part of the spectrum is this on?" A salesperson who lost a deal by testing a new pitch made an experiment. A salesperson who lost it by ignoring the qualification checklist made a process failure, which is usually a training or design problem. A salesperson who lost it by lying to the customer made a deviance failure, and that one is a disciplinary matter. Treating all three the same way, in either direction, is the mistake managers make most often.

Blameless postmortems: how engineering operationalised the idea

Software teams turned the principle into a routine. Google's Site Reliability Engineering handbook devotes a chapter to postmortem culture and states the rule plainly: "for a postmortem to be truly blameless, it must focus on identifying the contributing causes of the incident without indicting any individual or team for bad or inappropriate behavior". It assumes "that everyone involved in an incident had good intentions and did the right thing with the information they had", and it explains the cost of the alternative: "if a culture of finger pointing and shaming individuals or teams for doing the 'wrong' thing prevails, people will not bring issues to light for fear of punishment" (Google SRE, Postmortem Culture: Learning from Failure).

The broader framework comes from safety science. Sidney Dekker's work on just culture, drawn from aviation and healthcare, distinguishes retributive responses (who broke the rule and what do they deserve) from restorative ones (who was hurt, what do they need, whose obligation is it to meet that need), and argues that organisations which default to blame get less reporting and therefore less safety (Dekker, Just Culture). None of this is specific to software or aircraft. A blameless review works for a mis-sent invoice, a botched customer escalation or a bad hire.

A manager's playbook for the week after a mistake

1. Contain first, analyse second. Fix the immediate damage before asking how it happened. Analysis under pressure produces blame, not insight.
2. Write down the timeline with the person who made the mistake. Ask what they knew at each step and what they expected to happen. The goal is to reconstruct a reasonable decision made with incomplete information, which is what most mistakes are.
3. Place it on the spectrum. Experiment, complexity, process gap, skill gap, inattention or deviance. Say which one out loud. Only the last two involve the individual's conduct.
4. Fix the system, not just the person. If a process gap let the mistake through, the same gap will let the next person's through. Most $600,000 mistakes are $600,000 because nothing stopped them at $6,000.
5. Share the lesson without the name. A short write-up circulated to the team turns one person's expensive education into everyone's cheap one. That is the entire economic argument for keeping the employee.
6. Follow up in 30 days. Check that the process change stuck and that the person is not carrying the mistake around. People who are quietly punished, passed over or left out after an error learn the same lesson as people who are fired.

Hiring for people who learn from mistakes

The cheapest time to build a learning culture is before someone joins. Structured interviews can test for it directly. Ask candidates to describe a significant mistake they made, what they did in the first hour, what they changed afterwards and what they would tell a colleague facing the same situation. Candidates who describe the system, the timeline and the fix are showing you the behaviour Edmondson measured. Candidates who cannot name a mistake, or whose mistakes are always someone else's, are telling you something too.

Scoring those answers consistently across interviewers is where hiring teams usually slip. A shared interview evaluation form with a defined scale for "ownership and learning" helps interviewers score the same evidence the same way, and an applicant tracking system such as 100Hires keeps every interviewer's scores and notes on the candidate record so the debrief is about evidence rather than impressions. The same discipline applies once people are hired: a clear onboarding process is the first signal that the company expects questions and tolerates early errors, which is one reason onboarding matters so much for retention.

Back to the $600,000 question

So what would you do if an employee made a $600,000 mistake? The research answer is: find out where it sits on the spectrum, fix whatever let it grow to $600,000, publish the lesson, and reserve discipline for deliberate or reckless conduct. That is not soft management. It is the option with the lower cost, the higher reporting rate and, if Gallup's numbers hold, the better safety record. Watson may never have said the exact words, but he ran one of the most successful companies of the twentieth century on the idea, and the evidence has caught up with him.

For the wider picture on stress, burnout and support at work, see our guide to mental health at work.

Sources

• Peter E. Greulich, Tom Watson Sr.: We Forgive Thoughtful Mistakes (on the origin of the story)
• Gallup, This Fixable Problem Costs U.S. Businesses $1 Trillion (replacement cost estimate)
• Gallup, How to Create a Culture of Psychological Safety
• Amy Edmondson, Psychological Safety and Learning Behavior in Work Teams, Administrative Science Quarterly, 1999 (Harvard DASH)
• Amy Edmondson, Strategies for Learning from Failure, Harvard Business Review, April 2011
• Google re:Work, Understand team effectiveness (Project Aristotle)
• American Psychological Association, Work in America 2024: Psychological Safety in the Changing Workplace
• Google SRE, Postmortem Culture: Learning from Failure
• Sidney Dekker, Just Culture: Restoring Trust and Accountability in Your Organization