What is a Decision Review?

Most firms review outcomes constantly and decisions never. The quarter is examined, the programme is retrospected, the troubled account is picked over, and through all of it one question goes unasked: was the decision that set this in motion a good decision when it was made?

A decision review is the retrospective that asks it. It examines a past decision on three axes: what was decided, what was knowable at the time, and what has happened since. Its subject is the reasoning, not the person. Its output is not a verdict but a lesson the next decision can inherit.

The reasoning, not the person

The discipline hangs on one distinction: judging against what was knowable then, not what is known now. Hindsight makes every missed signal look luminous, and a review that reads the past through the outcome will always find someone to have been obviously wrong. The honest questions are harder. Given the evidence available at the time, and its quality, was the reasoning sound? Were the right options even on the table? Was the dissent heard, and was it weighed or merely minuted?

This is why a decision review separates decision quality from outcome quality. A good decision can produce a bad outcome, because the world moved. A bad decision can produce a good outcome, because luck exists. Judging by results alone teaches exactly the wrong lessons: it punishes well-reasoned bets that failed and canonises reckless calls that happened to land. Poker players have a word for this, resulting, and a firm that results its way through its own history learns to be lucky, which is not learning.

A decision review judges the reasoning against what was knowable at the time, and it can only do that if the reasoning was written down while the decision was live.

Why blame produces defensive records

The moment reviews assign blame, people start writing for the tribunal. Rationales turn vague enough to be defensible from any direction. Assumptions stop being falsifiable. Estimates widen until they cannot be wrong. The real reasons retreat into hallway conversations and private messages, where no review will ever find them. The organisation ends up with a record built for defence rather than for learning, and the next review has nothing honest to read. Blame does not just poison the meeting; it poisons the archive.

The causality runs the other way too. Reviews go looking for a culprit precisely when the record is thin, because when nothing was written down, what were we thinking has nowhere to go but who was responsible. A rich record makes the systemic reading possible: not who got this wrong, but what class of evidence do we keep underweighting, and what would we change about how the next decision of this shape is made.

One concrete example

Clearly illustrative, with no customer implied. A firm declines to match a discounted renewal on a large account, and within the year a competitor has taken the work and grown it. The blame version of the review asks whose call that was, and its main product is reflex: the next three renewals are matched without analysis, at margins nobody believes in. The decision review reconstructs the decision as it stood. The margin facts were measured, and they were genuinely poor. The client’s expansion intent was on the record but only ever stated, never tested. On the evidence as weighed, declining was defensible. The lesson is therefore not about a person at all. It is about a class of evidence: client-stated growth intent was systematically underweighted because it was soft. The next renewal decision inherits that lesson, and inherits the record that taught it.

The decision review in the ONX vocabulary

A review is only as good as the record it reads. A decision audit trail preserves who decided, what was recommended, and what was overridden against what evidence. Immutable scenario runs preserve what was believed at the time, with each fact’s evidence state attached, so hindsight cannot quietly edit the past. Scored outcomes accumulate in an Outcomes Ledger, which sets what was recommended, what was decided and what happened side by side, decision after decision. And the loop closes the way decision intelligence intends: patterns are mined from scored outcomes only once enough evidence accumulates, and fed back into future recommendations. At that point the review stops being an event and becomes a property of the system: every decision arrives already carrying the lessons of the ones before it.