Errors as Process Signals: Why Repeat Mistakes Rarely Come Down to People

Why the same mistake keeps happening

A delivery goes to the customer's old address, for the third time this year. Each time it was someone else, and each time the same reaction followed: a serious talk, a request for more care, an email to everyone. The fourth time, the owner's patience runs out, and the only question left is who was to blame this time.

The question is understandable, but it leads to a dead end. When three different people make the same mistake, it is unlikely that all three were careless. It is more likely that the workflow invites the mistake: the new address is in an email but not in the system, or the system shows the billing address instead of the delivery address at dispatch. This essay pursues the question: why do mistakes that repeat rarely come down to individuals, and what follows for how they are handled? To answer it, we first introduce a thinking model from safety research, then examine what studies show about how companies deal with errors, weigh the objections, and finally draw the conclusions for the company.

A mistake that happens once may come down to a person. A mistake that happens three times comes down to the workflow.

The thinking model: the person approach, the system approach, and the Swiss cheese model

The British psychologist James Reason shaped research on human error for decades. In a widely cited article, Reason (2000) distinguishes two ways of looking at errors. The person approach sees the cause in the people who make a mistake: in inattention, forgetfulness, or lack of care. Its answers are reprimands, training, and if necessary sanctions. The system approach assumes that people are fallible and that errors occur even in the best organizations. It does not ask who failed but why the safeguards failed.

To illustrate this, Reason uses an image that has become known as the Swiss cheese model. Every protective layer of a workflow, such as an input screen, a check, or a handoff, is like a slice of cheese with holes. The holes are weak points, and they move, depending on the day, time pressure, and staffing. Damage only occurs when the holes of several slices line up for a moment. Reason distinguishes active failures, which people commit at the point where the damage occurs, from latent conditions that have been built into the workflow long before, such as an unclear screen, a missing handoff, or constant time pressure.

One aid for making the distinction is the so-called substitution test, which Reason (1997) takes up in his work on safety culture. You imagine that another person with comparable training and experience had been in the same place under the same conditions. Could they have made the same mistake? If the answer is yes, replacing or reprimanding the person helps little. The cause lies in the conditions.

What studies show about how companies deal with errors

Whether the system approach also pays off economically was studied by Van Dyck, Frese, Baer, and Sonnentag (2005). They describe an error management culture as an organization's norms and common practices in dealing with errors: talking about errors, sharing knowledge about errors, helping each other in error situations, and detecting and correcting errors quickly. In a first study with 65 Dutch organizations, such a culture was related both to the achievement of organizational goals and to an objective indicator of economic performance. A second study with 47 German organizations confirmed the relationship using change-in-profitability data.

The authors distinguish two strategies that complement each other. Error prevention tries to block wrong actions from the outset. Error management accepts that errors can never be entirely avoided and aims to keep their negative consequences small and to learn from them. Their criticism is directed at organizations that rely on prevention alone. Because errors can never be ruled out completely, a pure prevention strategy wastes the opportunity to learn from them.

Fischer, Frese, Mertins, and other authors (2018) transferred the idea to innovation. In 30 companies with 227 employees, an error management culture was related to the companies' innovativeness, measured using separate subgroups of the workforce to reduce bias from having the same respondents. This fits a finding described in the article on learning from bad decisions: Edmondson (1999) found in 51 work teams that psychological safety, the belief that errors can be raised without negative consequences, was related to learning behavior.

The objection: aren't there careless employees after all?

The obvious objection to the system approach is: they do exist, the colleagues who work sloppily. Whoever attributes every mistake to the workflow no longer holds anyone accountable and ends up rewarding the careless. This objection deserves a serious answer, because it is not wrong.

The system approach does not deny that people differ in how careful they are. It denies that this is the most common explanation for mistakes that repeat. The substitution test draws the line. If another capable person could have made the same mistake, the cause lies in the workflow. If someone deliberately ignored a clear, known, and sensible rule, it lies with the person. The article on bad decisions draws the same line: a mistake in an unclear workflow is an occasion for learning, a deliberate rule violation is a topic for a one-on-one conversation.

Research on work instructions supports this distinction too. In their study of 61 seafarers, Rajapakse and Emad (2025) explicitly oppose attributing accidents too quickly to human error, because deviations from procedures often arise from poorly prepared procedures and lack of training. And Zhang and Narayandas (2026) observed in a field experiment in customer service that even well-supported service agents could achieve least with repeat complaints, because the causes lay in the company's processes. I therefore consider the system approach the right first view and the person approach the exception that has to be justified.

How a company reads errors as signals

Whoever wants to treat errors as process signals first needs a place where they are collected. A simple error register is enough: date, what happened, at which point in the workflow, which conditions were involved. What matters is that no names need to be entered. The register serves the workflow, not the personnel file. After a few months it shows patterns that are not noticeable in individual cases.

The pattern in the worked example is typical, even though the numbers are freely chosen. The largest group concerns handoffs, and that is exactly where the article on interfaces between departments comes in. Mistakes that a screen or template invites can often be eliminated with a small change. Unclear rules belong in a better work instruction, as described in the article on how to write work instructions, or in a short checklist following the pattern from the article on checklists in the company.

1. Record without names

   Every error is noted with its place in the workflow and the conditions involved, not with the person's name.

2. Apply the substitution test

   For every recurring error, the team asks: could another capable person have made the same mistake at this point?

3. Find the slice

   Determine which protective layer failed: information, screen, rule, check, or capacity.

4. Change the workflow

   Change the condition so that the mistake is no longer possible this way or becomes visible immediately.

5. Check the effect

   Set a date on which you check whether the error has recurred since.

What this means for the owner's role

How a company deals with errors is decided not in the register but in the first minutes after a mistake. If the owner first asks who did it, the team learns to hide or downplay mistakes. If the owner first asks how it could have happened, the team learns to report mistakes early. Van Dyck and colleagues describe exactly these practices, talking about errors and detecting them quickly, as the core of an error management culture.

This does not mean giving up on standards. A company can set high demands for care and still read mistakes as a signal about the workflow. Both fit together because they concern different questions: the demand concerns behavior, the analysis concerns conditions. Whoever separates the two can be clear without becoming unfair. The essay on how to build a sales culture describes how this attitude shows in the way a team talks about customers and mistakes.

Conclusion: why do repeat mistakes rarely come down to people?

The question was
mistakes that repeat rarely come down to individuals and what follows from this. The answer lies in the repetition itself. When different people make the same mistake at the same point, this indicates a condition in the workflow that invites the mistake. The Swiss cheese model explains why such conditions go unnoticed for a long time: damage only occurs when several weak points coincide, and then the person at the end of the chain becomes visible, not the holes before. Studies on error management culture show that companies that talk openly about errors and fix them quickly do not fare worse economically but tend to do better. Personal responsibility remains, but it is the justified exception, not the first assumption. A mistake that repeats is free information about where the workflow is weak, and it would be uneconomical to ignore it.

How to get started

1. Take the mistake that has annoyed you most often in the last three months.
2. Determine how often it happened and to how many different people.
3. Apply the substitution test: could another capable person have made the same mistake at the same point?
4. Go through the workflow with those involved and identify the slice where the hole is.
5. Change this one condition and set a date on which you check whether the mistake has recurred.

If the mistake stays away afterward, you have achieved more than with any email to everyone. And your team has seen that a reported mistake leads to a better workflow and not to a culprit.