Discover
Right Kind of Wrong · Amy C. Edmondson
Management & Leadership · Free text

Right Kind of Wrong

Amy C. Edmondson

The Science of Failing Well

25 min
ListenRead

About the book

In Right Kind of Wrong, Amy C. Edmondson argues that neither “never fail” nor “fail fast” is a sufficient rule for doing good work. She offers a practical way to distinguish preventable failures in familiar tasks, complex failures produced by interacting conditions, and intelligent failures that can arise when people test thoughtful ideas in genuinely new territory. The book focuses on learning without glamorizing carelessness, replacing blame with clear-eyed inquiry, and building teams that can speak honestly about what went wrong.

Who this book is for

This summary is for managers, founders, project leaders, teachers, and individual contributors who need to learn from setbacks without lowering standards. It is especially useful for people working where some tasks are routine and others are uncertain, creative, or highly interdependent. Readers looking for permission to be reckless will not find it here; readers seeking a more humane and rigorous way to prevent avoidable errors, investigate surprises, and run responsible experiments will.

About the author

Amy C. Edmondson is a professor of leadership and management at Harvard Business School and a researcher of psychological safety, teamwork and organizational learning. She earned her doctoral, master’s and undergraduate degrees at Harvard University. Her books include The Fearless Organization and Teaming. Her work examines how people share information, learn across professional boundaries and improve performance in uncertain environments, connecting academic research with the practical challenges of leading teams.

Free text

Summary and key ideas

Book summary prepared by ReWio. Original book author: Amy C. Edmondson.

Introduction

Failure is a word that carries a remarkable amount of emotional weight. It can mean embarrassment, loss, disappointment, wasted effort, damaged trust, or real harm. Yet it can also be part of discovery. Because these meanings are so different, Amy C. Edmondson’s central claim in Right Kind of Wrong is simple but demanding: we need to get much better at distinguishing among failures. The familiar slogans do not do this job. A culture that treats every failure as unacceptable drives concealment, hesitation, and silence. People protect themselves rather than report risks early. But the opposite slogan—fail fast, fail often—can be just as misleading. It may turn carelessness into a badge of creativity, as if a bad outcome automatically proves that someone was brave or innovative. Edmondson argues that failing well is not about admiring failure. It is about pursuing worthwhile goals while reducing avoidable harm and extracting real learning when outcomes are disappointing. Her definition begins with results. A failure is an undesired outcome. That does not mean every error is a failure, or every failure comes from a single error. Nor does it mean every unsuccessful attempt deserves praise. A typo caught before publication may be an error but not a failure. A product launched carefully that customers do not want may be a failure even when the team acted intelligently. This distinction matters because correct diagnosis determines the correct response. The book’s framework sorts failures into three broad kinds: basic, complex, and intelligent. Basic failures occur in relatively known situations when a familiar process, rule, or skill is not properly applied. Complex failures emerge from multiple factors interacting in ways no one person completely controls. Intelligent failures arise when people venture into new territory with a thoughtful, bounded experiment and receive an unwanted but informative result. These categories are not excuses, personality labels, or moral rankings. They are practical lenses. They ask: How much is known about the task? How interdependent is the system? What could reasonably have been anticipated? What did we test, at what scale, and why? Answering these questions shifts the conversation from “Who failed?” to “What happened, what can we prevent, and what can we learn?”

Stop Treating All Failure Alike

The most useful feature of Edmondson’s framework is that it resists one-size-fits-all reactions. When leaders respond to every setback with the same speech about resilience, people receive an unclear message. A team may overlook a preventable safety check, struggle with a tangled handoff between five departments, and run a careful pilot that disproves a promising idea. These should not be managed as identical events. Basic failures call for consistency. If a process is known and repeatable, the organization should make success easier through training, accessible instructions, sensible design, checklists, peer review, and feedback. “Basic” does not mean trivial. A basic failure can be costly or dangerous. The label refers to how understandable the task is, not to the seriousness of the outcome. Repeatedly skipping a well-established step is not innovation. It is a signal to improve discipline, resources, clarity, or the process itself. Complex failures require a wider lens. In an interconnected environment, several individually understandable decisions can combine into an unwanted result. One team runs late, another works from an outdated assumption, a handoff is unclear, and a customer request changes at the wrong moment. Searching immediately for one culprit may feel satisfying, but it often makes the system less visible. A stronger response maps conditions, dependencies, timing, incentives, and weak signals. The aim is not to deny individual responsibility; it is to understand how the whole situation made the outcome more likely. Intelligent failures belong in a third category. They occur when the answer genuinely is not yet known. A team has a meaningful opportunity, a plausible hypothesis, and a method for testing it. It acts at the smallest practical scale, learns that the idea does not work as expected, and shares the lesson. The unwanted outcome is still unwanted. What makes it intelligent is the quality of the learning design, not the team’s optimism or the novelty of its vocabulary. Consider these original illustrations. A familiar operations team uses an established closing checklist and misses one required verification. That is likely a basic failure: repair the process and make reliable completion easier. A delivery arrives late because sales, inventory, packaging, and transport data failed to align. That is likely complex: investigate the chain rather than blame the final courier. A small business offers a new service to a few willing customers before making a large investment, then finds demand is weak. That may be an intelligent failure, provided the test was ethical, affordable, and capable of answering a real question. Classification is not an academic exercise. It tells us whether the next move is prevention, systemic investigation, or another better-designed experiment.

Prevent What Is Already Knowable

Edmondson pushes back against the romantic idea that good organizations embrace every mistake. In known work, excellence often means making fewer errors. A reliable routine does not limit human potential; it can protect time, attention, money, and trust for the areas where imagination is actually needed. Prevention begins with respect for human fallibility. People get distracted, misunderstand instructions, forget steps, and work under pressure. A robust system does not rely on heroic concentration alone. It builds useful prompts into the workflow, makes critical information easy to find, reduces unnecessary complexity, and allows someone else to catch a problem before it becomes consequential. Training matters, but training without a workable environment asks people to compensate endlessly for a bad design. Accountability belongs here, but it needs precision. Accountability is not humiliation, nor is it the automatic search for a person to punish. It means taking ownership of one’s contribution, acknowledging effects on others, making amends where possible, and participating in a credible prevention plan. If someone repeatedly ignores a known procedure after fair training and support, that calls for a direct performance conversation. If the procedure is confusing, inaccessible, or routinely impossible under actual workloads, leaders have accountability too. An original personal example makes the point. Suppose you are learning a new skill and repeatedly miss practice sessions because you rely on memory after an exhausting workday. Calling yourself lazy supplies little information. A better analysis asks what is known: you forget at that time, your materials are not ready, and no cue protects the commitment. Put the materials in sight, schedule a smaller session, and track it. This is not an experiment in the sense of gambling with a high-stakes outcome. It is a practical redesign of a familiar routine. The same logic applies to teams. After a missed deadline, a manager can ask: Which commitments were clear? What early signal was visible? Who needed information sooner? Which part of the work was routine, and which part was uncertain? The answers may reveal a basic failure hidden inside a complex project. That distinction prevents two equally unhelpful reactions: dismissing an avoidable mistake as unavoidable complexity, or blaming an individual for conditions they could not control. The standard is not perfection. It is a steady effort to make predictable work reliably safe and effective. That effort creates the capacity to take wiser risks elsewhere.

See Systems, Signals, and Near Misses

Complex failure is difficult because the story rarely fits inside one person’s actions. Modern work is full of interdependence: different specialties, changing priorities, software tools, suppliers, shifting customer needs, and incomplete information. In these settings, undesirable outcomes can emerge even when no one intended harm and many people acted reasonably from their limited view. Edmondson’s practical implication is to replace reflexive blame with disciplined curiosity. Start by reconstructing what people knew at the time, not what seems obvious after the result is known. Hindsight can make uncertainty disappear. Once an outcome is visible, it is easy to say someone should have predicted it. Fair inquiry asks what signals were available, how clear they were, whether people had a channel to raise concerns, and whether they could act on those concerns. Near misses are especially valuable. A near miss is an event that could have produced a worse outcome but did not. Because no major damage occurred, teams may rush past it. That is a mistake. A near miss can reveal a fragile handoff, ambiguous ownership, or a warning signal that happened to be caught in time. Treat it as low-cost information. Ask what prevented escalation and whether that protection was dependable or merely lucky. Imagine an original example involving interconnected delivery handoffs. A customer receives the right order despite three problems: stock data were delayed, packaging instructions changed, and a driver noticed an odd label before departure. Celebrating the delivery alone would miss the vulnerability. A learning-oriented review would identify the weak interfaces, preserve the driver’s intervention as a useful safeguard, and improve the information flow before luck is needed again. Complex inquiry should be rigorous, not vague. A team can map the sequence of events, identify decision points, compare written procedures with actual work, and ask what trade-offs people faced. It can also look for recurring patterns across incidents rather than treating each one as isolated. The goal is practical redesign: clearer handoffs, better escalation routes, fewer conflicting metrics, or more time for coordination. This approach does not mean no one is ever responsible. Deliberate misconduct, disregard for essential safeguards, or refusal to address repeated warnings require a different response. But accountability becomes more useful when it is paired with a serious account of the system. If people learn that reporting a concern merely makes them a target, the organization loses the very information it needs to prevent the next failure.

Design Experiments Worth Failing

Intelligent failure is the book’s most hopeful idea, but also its most easily distorted one. An unsuccessful effort is not intelligent merely because it was new, expensive, or described as innovative. Edmondson’s standard is stricter: enter genuinely unfamiliar territory, pursue a meaningful opportunity, form an informed hypothesis, test at the smallest useful scale, and learn clearly from the outcome. The first condition is novelty. If an organization already knows how to perform a task safely and reliably, deliberately ignoring that knowledge is not experimentation. The second is purpose. A test should answer a decision-relevant question, not merely keep a team busy. The third is preparation. You do not need certainty, but you should have a reason to expect a result and a way to recognize what the result means. Small scale is an ethical and economic discipline. It limits wasted resources and reduces the chance that others bear an unfair cost for your learning. Before testing, define boundaries: Who has consented? What is the budget? What is the time limit? What outcome would make us stop, adapt, or expand? These questions are vital whenever other people, scarce resources, or reputations are at stake. Take the original small-business illustration. Rather than building an entire new service line, the owner offers a limited version to a small group of willing customers. The owner states what is being tested, collects feedback, monitors cost and demand, and decides in advance what evidence would justify further investment. If customers do not value the offer, the pilot has failed in its immediate goal but may have succeeded as learning. The owner avoids a much larger, less informative failure. Reflection completes the cycle. After an experiment, ask: What did we expect? What happened? What surprised us? Was the hypothesis wrong, the execution weak, the measure inadequate, or the context different? What will we do next? Learning is not automatic because experience is not self-interpreting. People can protect their ego, cherry-pick evidence, or turn an isolated result into a grand conclusion. This is also where persistence must be separated from stubborn repetition. Continuing through difficulty can be admirable when the evidence suggests the approach can improve. Persisting after repeated disconfirming evidence without changing the approach is often avoidance disguised as grit. A useful rule is to persist in the purpose while remaining flexible about the method. There are firm limits. No one should run casual experiments with safety-critical established procedures. Novelty does not excuse avoidable harm. High-stakes work requires stronger safeguards, expert review, and appropriate consent. The lesson is not to experiment everywhere. It is to experiment responsibly where uncertainty is real and learning can guide the next move.

Turn Defensiveness Into Curiosity

A difficult outcome threatens identity. We want to be competent, careful, and good people. When evidence challenges that image, defensiveness arrives quickly: denial, excuses, counterattack, silence, or a rush to move on. Edmondson’s work takes this human reaction seriously. Learning from failure is not only an analytical problem. It is an emotional one. The first practice is to create a pause between the outcome and the verdict. Instead of immediately deciding that someone was careless or that you are incapable, describe the facts as plainly as possible. What was the intended result? What actually occurred? What decisions and conditions came before it? This does not remove emotion, but it gives emotion less control over the investigation. Self-awareness means noticing the story you are tempted to tell. Perhaps you are protecting your status, blaming circumstances, or interpreting one setback as proof that you should quit. Curiosity asks more useful questions: What am I not seeing? What information would change my mind? Whose perspective is missing? What part of this can I own without claiming responsibility for everything? For an individual, accountability can include a clear apology, repairing consequences, and naming the changed behavior that will follow. For a manager reviewing a missed deadline—again, an original illustration—it can mean beginning with inquiry rather than accusation: “Help me understand what made this hard to surface earlier.” The manager can then distinguish a capability gap, a resource conflict, an unclear priority, and a preventable oversight. Each needs a different remedy. Curiosity should not become endless analysis. A good review ends in action. Decide what to retain, what to change, who owns the next step, and when to check whether the adjustment worked. Share lessons at the right level so the same insight does not remain trapped with one person or one team. The deeper message is compassionate realism. Human beings are fallible. Admitting this is not lowering expectations; it is the starting point for designing wiser habits, teams, and institutions. Shame narrows attention. Constructive accountability broadens it toward repair and improvement.

Speak Up While Standards Stay High

Psychological safety is an essential condition for learning from failure, but it is often misunderstood. It does not mean permanent comfort, automatic agreement, low standards, or freedom from consequences. It means that people believe they can raise work-relevant questions, concerns, ideas, and mistakes without being humiliated or punished for speaking up. That distinction is crucial. A team can be friendly yet unsafe if members avoid bad news. It can also be candid and psychologically safe while debating intensely, giving demanding feedback, and holding high performance standards. Edmondson’s position is that safety and accountability work together. Safety removes the interpersonal fear that keeps useful information hidden; accountability keeps attention on quality, commitments, and results. Leaders shape this environment through repeated small actions. First, frame the work honestly. If a task involves uncertainty, interdependence, or learning, say so. Explain why no one person can see the whole picture and why questions are necessary. Second, invite participation with genuine questions, especially from people closest to the work or least likely to speak. Third, respond productively when someone brings bad news. Thank them for raising it, ask for details, and focus on the problem before assigning conclusions. A defensive reaction teaches a powerful lesson. If a leader visibly punishes the messenger, others will calculate that silence is safer. A calm, interested response does not mean the issue is unimportant. It means the team can now understand it early enough to act. The eventual response may involve retraining, redesign, a difficult accountability conversation, or stopping an unsafe practice. Candor and consequences are compatible. Edmondson’s earlier research also helps explain why a team with more reported errors may not necessarily be making more underlying errors. A more open team may simply reveal more of what was already happening. Reporting data therefore need interpretation. Treating every increase in reports as proof of worse performance can discourage disclosure and make the organization less informed. The practical test is simple: when someone reports a mistake or raises an inconvenient concern, what happens next? If the first response is curiosity and appreciation for the information, people learn that truth is usable. If the first response is embarrassment or blame, they learn to hide it. Over time, that difference determines whether a group merely reacts to visible failures or learns before they grow.

Final Summary

Right Kind of Wrong offers a disciplined alternative to both fear of failure and enthusiasm for failure. Its key insight is that bad outcomes differ, and intelligent action starts by telling those differences apart. Prevent basic failures in familiar work through training, thoughtful systems, checks, and clear accountability. Investigate complex failures by examining interactions, handoffs, signals, and conditions rather than hunting for a single convenient culprit. Pursue intelligent failures only in genuinely new territory, with a meaningful hypothesis, a limited and ethical test, and explicit reflection afterward. The book also asks for a change in posture. We need enough humility to admit that we are fallible, enough courage to speak up early, and enough curiosity to examine evidence that threatens our preferred story. Accountability then becomes a route to repair and improvement, not a ritual of blame. For leaders, the task is concrete: clarify what is known and uncertain, set high standards and clear boundaries, invite dissent, respond well to bad news, and make learning visible in the next decision. For individuals, the task is to trade self-protection for honest reflection: own what is yours, understand the system around you, and alter your method when evidence calls for it. The right kind of wrong is not a failure to celebrate. It is an outcome from which a careful person or team can learn something real—while working steadily to prevent the kinds of failure that never needed to happen.

Related books