A Good Decision Can Still End Badly: How to Review Choices Without Outcome Bias

You researched the options, considered the risks and made a reasonable choice. Then the outcome went badly. Was the decision wrong?

Perhaps—but the bad result alone cannot answer that question. A careful choice can meet bad luck, just as a careless choice can be rescued by good luck. When we judge the process entirely through the result, we learn the wrong lesson from both.

This distinction matters in business, leadership, parenting, careers and ordinary life. If success excuses weak reasoning, reckless habits can become policy. If failure automatically condemns a sound process, people become afraid to take sensible risks.

Outcome quality and decision quality are different

An outcome is what happened. Decision quality concerns how the choice was made using the information, options, values and uncertainty available at the time.

In five experiments published in 1988, psychologists Jonathan Baron and John Hershey gave participants descriptions of decisions made under uncertainty, followed by favourable or unfavourable results. Even when participants understood that the decision-maker had possessed the same relevant information, they rated the thinking and competence more positively after good outcomes than after bad ones. The researchers called this outcome bias.

The bias is understandable. Once the result is visible, it becomes vivid evidence. The risks that did not occur fade, and the path that happened can feel as though it was always the likely path. But the decision-maker did not have the ending when the choice was made.

A simple example: the careful supplier choice that failed

This example is fictional and created for explanation.

A small retailer compares two suppliers. Supplier A is cheaper but has inconsistent delivery records. Supplier B costs more, provides audited quality documents and has a strong record of on-time fulfilment. The retailer chooses B.

Three months later, an unexpected port closure delays Supplier B’s shipment. Supplier A happens to have stock and delivers quickly to its customers. Looking only at the outcome, choosing B appears foolish.

But a useful review asks different questions. Was the port disruption reasonably foreseeable? Did the retailer assess concentration risk? Could it have kept a small secondary supplier? Was the quality premium worth the known benefits? The failure may expose a missing contingency plan without proving that the original comparison was irrational.

Do not use “process” to excuse every failure

Separating process from outcome does not mean outcomes are irrelevant. A process that repeatedly produces poor results needs scrutiny. Nor should “bad luck” become a convenient label for risks that were ignored, hidden or never investigated.

The goal is a two-part review:

• Ex ante: Was the decision reasonable before the result was known?
• Ex post: What does the result reveal about the model, assumptions, execution or environment?

A good outcome cannot repair an irresponsible process. A bad outcome cannot, by itself, prove that the process was irresponsible. Both pieces of evidence matter, but they answer different questions.

The six-part decision-quality review

1. Reconstruct the information boundary

List what was genuinely known at the moment of choice. Separate it from information learned later. Ask what a reasonable person could have discovered with proportionate effort—not what now looks obvious.

This protects the review from hindsight. It also exposes negligence: if important information was available, affordable and relevant but deliberately ignored, that belongs in the assessment.

2. Check whether the real objective was clear

A choice can produce a visible success while failing the actual purpose. A campaign may generate clicks but lose trust. A cheap contract may save money this month while increasing service failures. A promotion may raise status while destroying the flexibility the person valued.

Write the objective as an outcome and a constraint: “Increase qualified enquiries without exceeding the budget or making claims we cannot support.” This is stronger than “make the campaign go viral”.

3. Examine the options—not just the chosen one

Was there a genuine choice? Were alternatives compared on the same criteria? Did the team include “do nothing”, “delay”, “test first” or “combine two options”?

The UK Government’s 2026 Green Book stresses presenting decision-makers with genuine choices rather than a single predetermined answer, and explicitly addresses risk, uncertainty and optimism bias. The principle travels well beyond government: a decision process is weak when it disguises advocacy as comparison.

4. Inspect assumptions and probabilities

What had to be true for the decision to work? Which assumptions were supported by evidence, and which were estimates? Did the plan consider a range of plausible outcomes or only the most attractive story?

A probability is not a promise. If an event with a 20 per cent chance occurs, its occurrence does not prove the estimate was wrong. But if supposedly rare events keep occurring, the estimate or underlying model may need revision.

5. Separate execution from selection

Sometimes the option was sensible but implementation was poor. A business chooses the right software but fails to train staff. A family agrees a helpful routine but never makes the cue visible. A project has sound goals but no owner, budget or review date.

Do not blame the choice for an execution failure—or praise the strategy when extraordinary effort merely rescued weak planning.

6. Compare with a counterfactual

Ask what might reasonably have happened under another option. This is not fantasy storytelling; it is disciplined comparison.

HM Treasury’s updated Magenta Book distinguishes observing an outcome from attributing that outcome to an intervention. Its evaluation guidance uses counterfactuals—credible estimates of what would have happened without the intervention—to assess contribution and causality. In daily decisions, we rarely have a perfect control group, but the question still improves thinking: “Would this result probably have happened anyway?”

Keep a decision receipt before the result arrives

The best time to protect against outcome bias is before you know the outcome. Create a short decision receipt containing:

• the decision and date;
• the objective and non-negotiable constraints;
• the options considered;
• the evidence available and its limitations;
• the main assumptions;
• the risks, likelihoods and possible impacts;
• what would make you reverse or revise the decision;
• the review date.

This record does not need to be long. Its value is temporal: it preserves what you believed before the ending rewrote the story in your memory.

Use four labels after the outcome

After reviewing the result, place the decision in one of four boxes:

• Good process, good outcome: preserve the method, but do not assume success was entirely skill.
• Good process, bad outcome: retain what was sound; improve resilience, assumptions or contingencies.
• Weak process, good outcome: celebrate the benefit, but fix the method before luck becomes confidence.
• Weak process, bad outcome: identify the preventable failures without turning the review into personal humiliation.

This grid prevents two dangerous organisational habits: rewarding recklessness because it worked once, and punishing careful experimentation because uncertainty produced a loss.

Moral luck: why outcomes still affect responsibility

Philosophy complicates the picture further. Two people may act with equal carelessness, yet only one causes serious harm because circumstances differ. We naturally judge them differently, even when the difference includes luck. This tension is often discussed as moral luck.

The practical answer is not to ignore consequences. Harm must be repaired, affected people supported and systems changed. But accountability should distinguish intention, reasonable foresight, control, process and consequence. Otherwise, punishment becomes a substitute for learning.

How leaders can review decisions without creating fear

1. Review the decision receipt before discussing who was responsible.
2. Ask which facts were available then and which emerged later.
3. Separate option selection, execution and external events.
4. Name both the successful safeguards and the missed signals.
5. Turn lessons into a changed checklist, trigger, limit or test.
6. Do not demand certainty that no one possessed.

The UK Government’s updated Test and Learn guidance advocates starting small, testing critical assumptions and adapting from evidence before scaling. That is useful management philosophy: reduce the cost of being wrong, make learning observable and avoid converting every experiment into a referendum on someone’s competence.

Judge the choice at the point where it was made

A mature review holds two thoughts at once: results matter, and results are not the whole story.

When a decision succeeds, ask whether the process deserves repeating. When it fails, ask whether the reasoning was sound, the uncertainty was respected and the execution matched the plan. Then update what genuinely needs updating.

You cannot make uncertainty disappear. You can make your decisions more honest, your reviews more fair and your learning less dependent on luck.

---

Sources and further reading

• Baron and Hershey: Outcome bias in decision evaluation
• HM Treasury: The Green Book 2026
• UK Government: Magenta Book evaluation guidance, updated 15 May 2026
• UK Government: Test and Learn guidance

Explore more MaryChuks philosophy articles and the related guide A Scenario Is Not a Prediction.

Featured image disclosure: the image accompanying this article is an original conceptual AI illustration. It does not depict a real person or documented decision.