Failure, and what it is reasonable to learn from it

Failure is widely described as a teacher. Whether people actually learn from it is less certain, and the research suggests several reasons why they frequently do not.

Why learning from failure is hard

It threatens self-concept, which prompts defensive processing rather than analysis.

Studies have found that people attend less to information about their own failures than about their successes, and remember it less accurately.

Attribution runs the wrong way. The self-serving pattern attributes success to internal factors and failure to external ones.

Which means the lesson extracted is frequently that circumstances were unfavourable, and no adjustment follows.

The inverse pattern — attributing all failure to personal inadequacy — is equally uninformative and considerably more distressing.

Outcome and process are conflated. A good decision can produce a bad outcome and a bad decision a good one, particularly where chance is involved.

Judging the decision by the outcome — outcome bias — produces the wrong lesson in both directions, and it is very difficult to resist.

The distinction that helps most

Separating decision quality from outcome quality.

The useful question after a failure is not whether the outcome was bad but whether the decision was reasonable given what was knowable at the time.

Where the decision was sound and the outcome poor, the lesson is about variance rather than about judgement — and changing the approach would make future decisions worse.

Where the decision was poor and the outcome happened to be good, there is a lesson that the outcome conceals entirely.

This framing comes originally from decision analysis and from poker, where the distinction between a good bet and a winning hand is unavoidable.

Types of failure

A taxonomy that is genuinely useful, drawn from organisational research.

Preventable failure. A known process was not followed. The lesson is about execution and the response is straightforward.

Complexity-related failure. Multiple things interacted in an unanticipated way in a system too complex to predict fully.

The lesson is rarely about individual competence and is usually about detection, redundancy or reducing coupling.

Intelligent failure. A deliberate experiment in uncertain territory that did not work.

This is the category that produces information and it should not be treated as failure in the blameworthy sense at all.

Organisations that treat all three identically get less of the third, which is the one they need.

Doing a useful review

Write it down close to the event, before memory reconstructs it into a story.

Hindsight bias is powerful and fast — outcomes come to seem predictable almost immediately, which corrupts the account of what was actually known.

Record what you believed at the time, not what you now think was obvious.

Keeping a decision journal — writing predictions and reasoning before outcomes are known — is the only reliable defence against hindsight bias, and it is a genuinely powerful practice.

Ask what you would need to have known. This distinguishes a knowledge failure from a judgement failure from bad luck.

Look for the systemic version. If a person made an error, ask what made the error easy — which usually produces a better fix than resolving to be more careful.

Identify one change, specific enough to act on. A review producing general resolutions produces nothing.

The overcorrection risk

Underdiscussed.

A single vivid failure produces disproportionate caution, because vivid events are weighted more heavily than their frequency warrants.

Which means the lesson from one bad experience is frequently too broad: avoiding a whole category of action because one instance went wrong.

Asking how many times this approach has worked, against how many times it has not, is a corrective that people rarely apply to themselves.

The limit of the framing

Worth saying plainly.

Not every failure contains a lesson. Some things fail because of factors nobody controlled, and the search for a lesson in those cases produces false ones and unwarranted self-blame.

Sometimes the accurate conclusion is that a reasonable attempt did not work, that nothing would have been done differently, and that the appropriate response is to try again rather than to change anything.

Accepting that is harder than extracting a lesson, and it is frequently more accurate.

General information about wellbeing, not psychological advice. Persistent distress following a setback warrants support from a qualified professional.