Learning From Setbacks: When Failure Becomes Useful Information

People often say that setbacks make us stronger. The phrase is appealing because it transforms disappointment into a promise: if something went wrong, eventually it must become useful. Research supports a more conditional conclusion. Difficulty can reveal information that success hides, but a setback does not automatically produce learning, resilience or better performance.

A failed exam can expose weak understanding. A rejected application can reveal poor positioning. An over-budget project can uncover unrealistic assumptions. A lost client can show that a sales process depended too heavily on one relationship. But setbacks can also exhaust people, reduce opportunities or result from circumstances over which they had little control. A person who experiences discrimination, serious illness or an economic shock does not become responsible for proving that the experience was somehow beneficial.

The useful question is therefore not whether setbacks are “good”. It is what information a setback contains and whether the person or organisation has the feedback, support and opportunity needed to use it.

That distinction matters because failure itself can interfere with learning. Experiments by Lauren Eskreis-Winkler and Ayelet Fishbach found that participants sometimes learned less from personal failure feedback than from success, even when failure contained enough information to identify the correct answer. The researchers found evidence that failure could threaten the ego and lead people to disengage from the information.

A setback becomes valuable only when something changes because of it: a belief becomes more accurate, a skill gap becomes visible, a process improves, a decision rule changes or a future attempt is redesigned.

Setbacks reveal what success can hide

Success provides evidence that something worked under particular conditions. It does not necessarily reveal why.

A project may meet its deadline because one employee quietly works unsustainable hours. A student may pass an examination by recognising familiar questions while still lacking deep conceptual understanding. A sales team may exceed its target during unusually strong market demand and conclude that its process is better than it really is. A business may grow despite weak controls because favourable conditions temporarily cover the weaknesses.

As long as the desired outcome continues, there may be little pressure to examine these hidden dependencies.

A setback interrupts that comfort.

When the project finally misses a deadline, the organisation may discover that critical knowledge sits with one person. When the student encounters a test requiring transfer rather than recognition, the limits of the previous study method become visible. When market conditions weaken, a sales process that depended on incoming demand may stop working.

The useful information is not simply that the outcome was bad. The setback exposes assumptions that had previously escaped testing.

This makes post-setback analysis different from self-criticism. The aim is not to ask, “Who failed?” before understanding what happened. It is to ask which assumptions the previous process depended on and which of them reality has now contradicted.

Was the schedule realistic? Did the plan assume that a supplier would never be late? Did the student confuse familiarity with recall? Did the manager assume technical expertise would automatically translate into leadership skill? Did the business interpret temporary market conditions as permanent demand?

These questions convert disappointment into diagnosis.

A setback creates information when reality differs from prediction

Learning often begins when an expected outcome and an observed outcome do not match.

You expected an interview to go well, but the panel repeatedly asked for examples you had not prepared. You expected an integration project to take three weeks, but dependency problems doubled the schedule. You expected a new role to suit you, but the everyday work turned out to be very different from what attracted you to it.

The gap between prediction and reality tells you that some part of your model needs revision.

The relevant model might concern the task: “I underestimated how much coordination this project requires.”

It might concern your current ability: “I understand the technical work, but I cannot yet explain it clearly under questioning.”

It might concern the environment: “This organisation rewards speed much more heavily than depth.”

Or it may concern the goal itself: “I was attracted to the status associated with this career more than to the work it requires.”

The value of the setback depends on identifying the right model. If the diagnosis is wrong, the lesson can make future decisions worse.

Someone rejected from one competitive job might conclude that they need an entirely different career when the actual issue was simply that another candidate had more relevant experience. A business whose product fails during an economic downturn might redesign the product even though demand was primarily affected by the wider market. Conversely, blaming every poor outcome on bad luck protects an inaccurate model from ever being corrected.

A useful review therefore separates observation from inference. What happened? What did you expect? Which assumption did the result contradict? What evidence supports the proposed explanation?

That is more demanding than saying “learn from your mistakes”, but it produces much more usable information.

Productive failure does not mean simply letting people fail

Education research provides a useful warning against turning failure into a slogan.

The concept of Productive Failure, associated particularly with researcher Manu Kapur, does not mean that students learn best when teachers simply leave them to struggle. It refers to a structured instructional design in which learners first attempt a complex, unfamiliar problem and generate possible representations or solution methods before receiving formal instruction. Their initial attempts often fail to produce the accepted solution, but those attempts can prepare them to understand the later explanation more deeply.

The sequence matters.

Learners first explore the problem. Then instruction helps compare their attempts, identify critical features and consolidate them into more formal concepts or methods. Failure is therefore one component of a larger learning design, not the educational objective.

Recent research continues to examine when this sequence works and which learners need more support. A 2025 study comparing Productive Failure with a related vicarious-failure design found that learner characteristics and metacognitive processes matter, with lower-prior-knowledge or lower-self-concept learners potentially requiring additional guidance.

This qualification is important beyond classrooms.

Managers cannot assume that unsupported struggle automatically develops employees. Schools should not mistake lack of instruction for productive difficulty. Organisations should not deliberately create avoidable failures in the belief that adversity itself builds competence.

Useful struggle normally requires boundaries, feedback and a later opportunity to understand what happened.

Errors can strengthen learning when correction follows

A related body of research examines errorful learning: situations in which learners attempt to retrieve or generate an answer, make an error and then receive corrective feedback.

Janet Metcalfe's review of the research concluded that, in many learning contexts, generating errors followed by corrective feedback can improve later memory for the correct information. The correction matters because the learner has actively attempted the problem and exposed what was missing or mistaken.

Experiments have even found situations in which producing an incorrect answer and then receiving the correct one leads to stronger later memory than simply studying the correct information from the beginning.

But again, the lesson is not that errors are intrinsically useful.

Feedback is crucial. Research on error correction shows that simply telling someone that an answer is wrong may be insufficient; effective correction requires the learner to notice the error and receive information that replaces or repairs it.

This distinction explains why repeatedly making the same mistake is not evidence of useful learning. An error becomes informative when the person can compare expectation with outcome, understand the correction and update future behaviour.

Workplace research points in a similar direction. A meta-analysis of 24 studies involving 2,183 participants found positive effects for error-management training, an approach that combines active exploration with explicit encouragement to encounter and learn from errors. The strongest effects appeared when people later needed to transfer what they had learned to tasks different from those used during training.

The important mechanism is therefore not pain or embarrassment. It is active problem-solving combined with usable corrective information.

Setbacks can improve calibration and future decisions

Another benefit of setbacks is improved calibration: developing a more accurate understanding of what you know, what you can do and where your limits currently lie.

Before performance is tested under realistic conditions, self-assessment can be misleading. Familiarity can feel like knowledge. Experience can feel like expertise. Confidence can be based on having completed a task many times rather than having performed it under demanding conditions.

A difficult test changes that.

A student may discover that rereading produced familiarity but not unaided recall. A newly promoted manager may discover that technical competence does not automatically create delegation or feedback skills. A freelancer may discover that doing excellent work and managing a client relationship are separate capabilities.

Disappointing evidence can still be valuable because it allows practice to become more specific.

It may also expose errors in decision-making rather than execution.

After an over-budget project, a manager might realise that every proposal was evaluated using optimistic assumptions and begin requiring downside scenarios. After choosing a course that proved unsuitable, a student might test future options through conversations with people doing the work, short projects or internships before making another large commitment.

The resulting lesson is not “be more cautious”. It is more precise: gather better evidence before making the next prediction.

This is one of the most transferable benefits of a setback. A person may never face exactly the same situation again, but the decision process can still improve. Better pre-mortems, pilots, checklists, reference checks, scenario planning or second opinions can prevent the same underlying error from appearing in a different form.

A lesson has become useful when it changes a future decision rule.

Sometimes the system needs to change, not the individual

Setbacks are often discussed as tests of personal resilience. In organisations, that framing can hide the more important lesson.

Suppose a deadline is missed because every important approval depends on one senior employee who was unexpectedly unavailable. Telling the project manager to become more resilient does not solve the vulnerability. The approval process needs redesign.

If employees repeatedly enter incorrect data because instructions are ambiguous, repeated retraining may be less effective than improving the interface or documentation. If teams routinely work unsustainable hours before every product launch, the organisation may have a capacity, scheduling or prioritisation problem rather than a motivation problem.

A useful post-project review therefore examines both individual actions and system conditions.

What made the error possible? What allowed it to remain undetected? Which safeguards were absent? Did incentives encourage risky behaviour? Was the process dependent on one person? Was relevant information available when the decision was made?

This approach does not eliminate accountability. People still make avoidable mistakes and sometimes ignore clear procedures. But individual blame can become analytically convenient because it closes the investigation quickly.

System-level learning asks a harder question: if another competent person were placed in the same conditions, how likely would the problem be to recur?

If the answer is “quite likely”, changing the person without changing the system leaves much of the risk intact.

Failure can also make learning harder

The idea that difficulty creates insight should not obscure the emotional effects of failure.

Eskreis-Winkler and Fishbach's research is important precisely because it demonstrates that informative feedback may still be ignored when it threatens how people see themselves. Participants in their experiments sometimes learned less from their own failures while learning normally from another person's failure, where the ego threat was reduced.

This helps explain why vague or humiliating criticism can produce defensive reactions rather than improvement.

“You are not leadership material” directs attention toward identity. “During the meeting, three decisions remained unresolved because responsibilities were not assigned” directs attention toward observable behaviour.

The second statement can still be uncomfortable, but it offers something that can be investigated and changed.

Psychological safety is relevant here, particularly in organisations. It should not mean removing standards, consequences or difficult feedback. It means creating conditions in which people can identify errors, admit uncertainty and report emerging problems without unnecessary humiliation or fear of disproportionate punishment.

When people hide mistakes, information arrives later. When they defend themselves rather than examine what happened, learning becomes more difficult. When asking for help is treated as incompetence, problems may grow until they become expensive failures.

A learning environment therefore needs both accountability and enough safety for accurate information to surface.

Some setbacks are primarily harm, not lessons

There are limits to any attempt to frame adversity as useful.

A discriminatory rejection, workplace abuse, serious injury, financial crisis, bereavement or family emergency may eventually change how someone understands life or makes decisions. That does not mean the event was beneficial or necessary.

Growth after adversity and the value of adversity are different claims.

People sometimes emerge from difficult experiences with new skills, priorities or relationships. But the possibility of growth should not be used to excuse preventable harm or pressure people into producing an inspiring story.

This distinction is particularly important in workplaces and education. An organisation should not defend poor management by saying employees will become resilient. A teacher should not equate unnecessary confusion with intellectual challenge. A system that creates avoidable harm does not become good because some people manage to adapt to it.

Nor does every setback contain a profound lesson.

Sometimes the evidence is simply that an unpredictable event occurred. Sometimes another candidate was a better fit. Sometimes the market changed. Sometimes a random failure happened despite a reasonable process.

Learning includes recognising when not to overreact.

If every negative result leads to a complete change of strategy, decisions become just as poorly calibrated as they would be if failure were ignored.

What makes a setback more likely to become useful?

Setbacks become more informative when several conditions are present.

First, there needs to be specific evidence about what happened. “The project failed” is less useful than identifying where costs exceeded assumptions, which dependency caused the delay and when the warning signs first appeared.

Second, people need corrective feedback. Knowing that something was wrong is often insufficient; the learner needs information that helps close the gap.

Third, there must be enough practical and psychological capacity to analyse the event. Someone still managing an immediate crisis may need recovery and stabilisation before detailed reflection becomes useful.

Fourth, learning needs a future test. If a person identifies a lesson but never encounters another situation in which the revised strategy can be tested, it remains largely theoretical.

Fifth, the response should match the scale of the problem. A small skill gap may require practice. A missing technical capability may require formal instruction. A structural process problem may require organisational redesign rather than coaching one individual.

Finally, a review should consider both personal actions and environmental causes. This prevents the two common extremes of assuming everything was personally controllable or assuming nothing can be changed.

These conditions explain why two people can experience similar setbacks and learn very different amounts from them. The event is only the starting point. What follows determines much of its educational value.

Use a setback to update one model

After the immediate consequences are contained, one practical approach is to identify the most important model that the setback has challenged.

Ask four questions:

• What did I expect to happen?

• What actually happened?

• Which assumption best explains the gap?

• If my revised explanation is correct, what should I predict or do differently next time?

The final question matters most.

Suppose someone concludes after a poor interview, “I need to communicate better.” That may be true, but it is too vague to test.

A more useful model might be: “When asked for evidence under pressure, I rely on general statements because I have not prepared specific examples.” The new action follows directly: prepare a small bank of evidence-based examples and test retrieval in mock interviews.

A project team might change “we need better planning” into “our schedules repeatedly ignore external dependency risk”. The new prediction becomes testable: projects using explicit dependency buffers and escalation triggers should experience fewer surprise delays.

If the supposed lesson creates no different prediction, decision or behaviour, it probably remains too abstract.

Setbacks are data with a cost

The most balanced view of setbacks is neither to hide them as evidence of inadequacy nor celebrate them as gifts.

They are events that carry costs and may also contain information.

Success tells you that a process produced the desired result under the conditions that existed. A setback tells you that some part of the expected relationship between action and outcome did not hold. Neither result explains itself.

Learning requires interpretation.

Sometimes the useful response is more practice. Sometimes it is better feedback. Sometimes it is a redesigned process, a different decision rule or a change in environment. Occasionally, the evidence suggests that the original goal itself deserves reconsideration.

And sometimes the event was simply harmful, unlucky or outside reasonable control.

The purpose of learning from setbacks is not to force every disappointment into a growth story. It is to extract information where information genuinely exists and use it to make the next model, process or decision more accurate.

A setback becomes useful when it changes what you know—and that new knowledge changes what you do.