Learning From Bad Decisions Without Looking for Someone to Blame

Why do businesses learn so little from their mistakes?

A large proposal was lost even though everyone was sure it would work. A new salesperson quit after half a year. A price increase cost two long-standing customers. In almost every business, such events are followed by the same pattern: brief excitement, an explanation that usually concerns one person, and then daily business goes on.

The result is that the next similar mistake is very likely to happen again. Not because the people involved are bad, but because nobody understood how it came about. The explanation “He misjudged it” is convenient. But it does not answer why an experienced person arrived at that judgment and whether someone else in their place would not have decided the same way.

That raises one question: how does a business learn from bad decisions instead of just finding someone to blame? To answer it, we first distinguish a bad decision from bad luck, then show why the question of blame blocks learning and what debriefs achieve, and finally introduce the Learning Review and how its findings become rules. It builds on the article Deciding Under Uncertainty, which describes how to tell a good decision from a good outcome.

Bad decision or bad luck: what actually went wrong?

Not every bad outcome is a bad decision. The article on deciding under uncertainty describes the mistake of judging a decision by its outcome alone. The author Annie Duke calls it resulting. A carefully weighed decision can turn out badly, and a careless one can turn out well. If you only look at the outcome, you learn the wrong thing in the first case and nothing in the second.

Before a business learns from a mistake, it therefore has to clarify what kind of mistake it was. Four causes can be distinguished, and each calls for a different response.

The fourth row matters. A business that changes its rules after every stroke of bad luck becomes more cautious and slower over time without getting better. That is exactly why a close look pays off: it protects you not only from repeating mistakes but also from drawing the wrong lessons from chance.

If you find a culprit after every mistake, you end up with many culprits and not a single better rule.

Why the blame question blocks learning

The question “Who did this?” has one advantage: it can be answered quickly. Its disadvantage is bigger. It changes what people tell. Anyone who fears that a name will be on the record at the end describes the sequence of events so that they look good. Important details are then missing, often not deliberately, but because memory of one's own role smooths itself out in the telling.

Research has coined a term for the opposite. Edmondson (1999) describes psychological safety as a team's shared belief that it is safe to take interpersonal risks in it, such as admitting a mistake, asking a question, or disagreeing. In her field study of 51 work teams in a manufacturing company, psychological safety was associated with learning behavior, meaning behaviors such as seeking feedback, talking about mistakes, and trying new things.

How a leader creates such safety was studied by Ngubane and Mbokota (2025) in South Africa. They interviewed 20 managers from 13 multinational consumer goods companies who led teams working partly in the office and partly from home. According to this study, the managers knew about the importance of psychological safety but did not treat it as a priority in practice. From this the authors develop a three-phase framework: first anchor a culture of safety across the company, backed by top management; then develop the leaders' necessary abilities, such as emotional intelligence and accountability; and finally reinforce the desired behavior through structures, standards, and recognition. The study is qualitative, small, and limited to one industry in one country. Its pointer is still useful to any owner: safety does not come from an announcement but from what actually happens after the next mistake.

What debriefs achieve

That structured debriefs work is well documented. In technical terms they are called debriefings or after-action reviews, a method the U.S. Army has used for decades and that is widespread in medicine and aviation. Tannenbaum and Cerasoli (2013) summarized the scattered research in a meta-analysis, a statistical analysis of many individual studies.

Their results from 46 samples with a total of 2,136 participants: debriefs improved performance compared with control groups by about 25 percent on average. The effect was similar for teams and individuals, in simulated and real settings, in medicine and outside it. Alignment between participants, focus, and purpose of the debrief proved especially important. Facilitation and a fixed structure also appeared to strengthen the effect. The authors explicitly call debriefs a relatively inexpensive and quick intervention.

One limitation lies in transferring the results. Many of the underlying studies come from training, medicine, and the military, where processes often repeat. In a small business's sales operation, situations are more varied. The basic idea still holds: anyone who reflects in a structured way after an event, discusses it, and sets a goal for next time gets better than someone who simply carries on.

What matters, however, is what happens after the debrief. Hicks (2026) describes, using disaster management as an example, that debriefs often produce detailed findings without anything changing durably afterwards. According to his analysis, the weak point lies in the transition from the documented finding to the verified change. A record that nobody acts on is not a learning experience.

The Learning Review: four questions in 45 minutes

The Learning Review is the form of debrief Owners Council recommends to small businesses. It takes 45 minutes, needs three to five participants, and follows four questions. It suits lost major orders, failed hires, pricing decisions that did not work out, and equally successes you want to repeat.

1. What did we expect?

   Read out the note written before the decision, ideally a Three-Sentence Note with expectation, rationale, and reversal signal. If there is none, everyone writes down for themselves what they expected at the time before anyone speaks.

2. What actually happened?

   Reconstruct the sequence in chronological order, facts only: who knew what when, what was done when. Judgments come only in the next step.

3. Where does the difference come from?

   Assign the cause to one of the four groups: wrong assumption, missing information, poor execution, or bad luck. With every answer, ask again: how did that come about?

4. What do we change, and who checks it when?

   Set at most two changes, each with an owner and a date on which you check whether it works. For bad luck the right answer is: change nothing.

The owner speaks last in the Learning Review. Whoever describes their view first sets the explanation in the room, and the others adapt. If the owner was involved in the decision, it is best to begin with their own share. That is the strongest signal that the point is learning and not blame.

How findings become rules

To keep the Learning Review from following the pattern Hicks describes, every change needs a fixed path into daily work. A simple learning log has proven useful: a list with date, occasion, cause, change, owner, and review date. On the review date you record whether the change worked.

If you keep such a log for a year, you see patterns that do not stand out in individual cases. The following worked example shows how the causes of twelve Learning Reviews might be distributed.

How to write down a new rule so that it is actually used in daily work is described in the article Documenting Processes With the One-Page Method. For lost orders in sales there is a dedicated approach tailored to this case, the win-loss analysis. And which numbers help test assumptions earlier in future is covered in the article Data-Driven Decisions.

What the Learning Review has to do with culture

A single Learning Review changes little. Only repetition shows the team that mistakes do not have to be hidden. After a few months the most important effect arrives: employees report problems earlier because they know they will be discussed and not punished. A mistake reported early is almost always cheaper than one that only surfaces at the customer.

That does not mean responsibility disappears. Anyone who repeatedly breaks clear rules or deliberately conceals mistakes must expect consequences. A learning culture distinguishes between a mistake that happens in an unclear process and a deliberate rule violation. The first is an occasion for learning, the second a topic for a one-on-one.

How the way people talk about customers and mistakes shapes sales culture is described in the essay Building a Sales Culture. How to give feedback on individual conversations without it coming across as reproach is shown in the guide Feedback in Sales Teams.

How to get started

Choose a decision from the past six months that did not turn out as hoped, ideally one of your own.

1. Invite two or three people involved to a 45-minute Learning Review and say beforehand that it is about the process, not about blame.
2. Ask everyone to write down before the meeting what they expected at the time.
3. Go through the four questions in order, and as the owner, speak last.
4. Set at most two changes with an owner and a review date.
5. Create a learning log and enter the first Learning Review. Also put the review date in your calendar.

On the review date you will see for the first time whether a change has worked. That is the moment a bad decision turns into a rule that benefits your business for good.

Back to the opening question, which was how a business learns from bad decisions instead of just finding someone to blame. The answer starts with setting the question of blame aside, because people who fear blame do not tell what really happened. According to the meta-analysis cited, well-run debriefs improve performance by around 25 percent on average. This only takes effect, however, when the review ends with a dated change that someone checks.