Mistakes are part of learning. So why do we punish them?
School taught many of us that mistakes cost points. Effective teams treat bounded mistakes as information, then turn that information into better decisions and systems.
“How can I ask a team to experiment when most of us were trained to treat every wrong answer as evidence that we failed?”
The short answer
Separate practice from performance.
A mistake made inside a bounded learning loop is information. A mistake concealed, repeated without reflection, or made recklessly in a high-consequence environment is a different problem. Healthy teams make that distinction explicit. They create room to be wrong while learning and clear standards for the moments when the work must be right.
Many of us learned a blunt lesson in school: correct answers earn points and mistakes lose them. A paper comes back with a score, the score becomes a grade, and the grade starts to feel like a judgment of the student rather than a snapshot of the work.
Grades can serve a real purpose. Schools need to assess understanding, communicate progress, and certify that certain standards have been met. The trouble begins when every attempt is treated like a final performance. If the first wrong answer permanently lowers the score, the rational strategy is not to explore. It is to avoid being seen while uncertain.
We carry that strategy into adulthood. We choose the project where we already know how to look competent. We soften bad news before it reaches a leader. We defend an early decision because changing our mind feels like admitting failure. We wait to share work until it looks finished, which means feedback arrives after it is cheap to use.
A system intended to measure learning can quietly teach us to perform certainty instead.
A mistake is unfinished information
This idea sharpened for me while reading Edward B. Burger and Michael Starbird’s The 5 Elements of Effective Thinking. One of their elements is fire: using mistakes to ignite insight. Their argument is not that failure is automatically noble. It is that an unsuccessful attempt can reveal assumptions, boundaries, and possibilities that a clean success leaves invisible.
That distinction matters. A mistake does not teach us merely because it happened. It becomes useful when we can observe it, explain it, and change the next attempt. Without that loop, failure is just cost.
| When a mistake is treated as a verdict | When a mistake is treated as a signal |
|---|---|
| The question — Who got this wrong? | The question — What did this reveal? |
| The response — Defend, conceal, or assign blame. | The response — Inspect, explain, and adjust. |
| The record — A permanent score attached to a person. | The record — Evidence that improves the next attempt. |
| The behavior it rewards — Choose safe work and project confidence. | The behavior it rewards — Test assumptions and surface uncertainty early. |
| The likely result — Late surprises and repeated hidden errors. | The likely result — Faster correction and stronger judgment. |
Learning requires a temporary gap between what we can do now and what we are trying to understand. That gap is where mistakes appear. Remove every possibility of being wrong and you have not created excellence. You have limited the work to what is already known.
The workplace often repeats the grading system
Leaders may say they want experimentation while their operating habits communicate the opposite. A forecast miss becomes a character judgment. A production incident becomes a search for the person who touched the last line of code. A changed recommendation is called inconsistency. A prototype is evaluated as though it were a launch.
People learn the real rules quickly. If bad news is punished, they delay it. If uncertainty is interpreted as weakness, they manufacture confidence. If every experiment must succeed, they stop running experiments and rename predictable work as innovation.
The cost is larger than morale. The organization loses information. Small problems stay private until they become expensive. Decisions remain unchallenged because disagreement feels risky. The company appears certain right up to the moment reality makes the correction for it.
Not every mistake belongs in the same category
“Mistakes are okay” is too vague to guide a team. The useful question is what kind of mistake occurred, under what conditions, and what the system should do next.
Exploratory mistake
A reasonable hypothesis failed inside a safe, reversible experiment. Preserve the learning and continue.
Execution mistake
A known process broke down. Fix the immediate issue, then improve the checklist, tool, handoff, or review that allowed it.
Repeated mistake
The same failure returned without evidence of correction. The issue is no longer the original error; it is the missing learning loop.
Reckless mistake
A known boundary was ignored where the consequence was serious. Accountability belongs here because the risk was not a necessary cost of learning.
This classification protects both sides of the idea. It prevents leaders from punishing thoughtful exploration, and it prevents “learning” from becoming a blanket excuse for carelessness.
Build a system that can learn
A learning culture is not created by telling people to be brave. It is created by changing the mechanics around the work.
- Label practice and performance. Say which work is exploratory and which must meet a fixed standard. A prototype, rehearsal, draft, and production release should not carry the same consequences.
- Bound the cost of being wrong. Use sandboxes, feature flags, small pilots, peer review, spending limits, and reversible decisions. Do not demand courage where basic safeguards would do more good.
- Write down the prediction. Before acting, record what you expect and why. Comparing the result with the prediction turns surprise into evidence instead of hindsight.
- Shorten the feedback loop. Review drafts early, test in small increments, and discuss incidents while the context is fresh. Feedback only improves learning when there is still time to act on it.
- Reward the correction. Notice when someone surfaces a risk early, changes their mind because the evidence changed, or improves the system after a miss. Otherwise the organization still rewards appearing right over becoming right.
Standards still matter
Some work has little room for trial and error in the live environment. Payroll, security controls, medical systems, financial reporting, and safety-critical operations need rigorous review and well-tested procedures. The answer is not to pretend people will never make mistakes. It is to move learning into safer places: simulations, staging environments, rehearsals, peer review, automated checks, and small reversible releases.
This is the same separation good education tries to make between formative feedback and summative assessment. Practice exists to reveal what the learner does not understand yet. Final evaluation determines whether the required standard has been reached. Confuse those two moments and people protect the score. Separate them and people can use feedback to improve before the score matters.
How this connects to the work I do
Architecture and delivery depend on the quality of a company’s learning loops. Roadmaps contain assumptions. Designs expose tradeoffs. Estimates are forecasts, not promises from an oracle. Incidents reveal where a system and its safeguards disagreed with reality.
I help teams make those loops explicit: smaller decisions, clearer hypotheses, safer release paths, earlier technical review, and incident practices that improve the system instead of merely locating blame. The goal is not a culture with lower standards. It is a culture that gets better evidence sooner and turns it into better execution.
Do not punish people for discovering that an assumption was wrong. Hold them accountable for what they do with that discovery.