A decision can be good and still turn out badly

Two things that get collapsed into one

The quality of a decision and the quality of its outcome are separate properties, and they come apart whenever chance is involved — which is to say, in almost everything that matters. A decision is good if it made sensible use of the information available at the time. An outcome is good if things worked out. The link between them is real but loose, and it gets looser as uncertainty rises.

Card and dice players have a word for the error of judging the first by the second, and they need one because the gap is visible to them every night. Most fields do not get that feedback, so the confusion persists undisturbed for years.

The four-cell picture

Lay it out and there are four possibilities. A good decision with a good outcome, which teaches you nothing you did not already believe. A bad decision with a bad outcome, which is painful and instructive. A good decision with a bad outcome, which feels like a mistake and is not. And a bad decision with a good outcome, which is the genuinely dangerous cell, because it rewards the behaviour and nothing signals that anything went wrong.

That last case is where most bad habits are learned. A gamble that paid off, a shortcut that happened not to be caught, a plan with no contingency that did not need one — each of these produces a positive result and reinforces a process that will eventually fail. The reinforcement is not weakened by the fact that it was undeserved.

Why the confusion is so hard to shake

Outcomes are visible, dated and unambiguous. Decision quality is an assessment of a counterfactual, requires reconstructing what was known at the time, and can be argued about. Any evaluation system will drift towards the thing that is easy to measure, and that is true of individuals reviewing their own history as much as of institutions reviewing their staff.

Language does not help either. We use the same word for both things, so a good decision and a good result are described identically and the distinction has to be carried by context that is usually absent. When someone says a choice was a mistake, they may mean the reasoning was faulty or they may mean it did not work out, and the two claims call for completely different responses. Most disagreements about past decisions turn out, on inspection, to be arguments in which each side is discussing a different one of these.

Hindsight makes it worse. Once the outcome is known, the reasoning that led to a bad result looks obviously flawed, because the mind has already rearranged the evidence into a path leading to what happened. Separating the two requires the contemporaneous record, and most of the time nobody kept one.

How to evaluate a decision properly

Reconstruct the decision from what was available at the time, without letting the outcome in. What options were on the table, what was known about each, what was uncertain, what could have been found out cheaply and was not? That last question is where most genuine criticism lives, because failing to gather easily available information is a process failure regardless of how things turned out.

Then ask whether the reasoning would still look sound if the result had been different. A decision that only looks defensible because it worked was not defensible. A decision that looks foolish only because it failed probably was not foolish.

This is not an argument for excusing every failure. It is an argument for locating the fault correctly, because the wrong diagnosis produces the wrong correction, and an organisation that punishes bad luck will get people who take no risks and hide the ones they take.

Sample size is the awkward part

Over a single decision, luck dominates and nothing can be inferred. Over many decisions of a similar kind, the process starts to show through, because chance averages out while systematic error does not. This is why evaluating a decision-maker on one result is close to meaningless, and why evaluating them on a hundred results is quite informative.

The uncomfortable implication is that in domains where you only make a few big decisions in a lifetime — a career, a major move, a large commitment — you will never accumulate enough cases to evaluate your own judgement from outcomes. The only available check is the quality of the process, examined at the time, which is precisely the thing outcome-based thinking teaches people to ignore.