What is a Decision Log?

Organisations minute their meetings, archive their contracts, version their code and file every invoice. The one artefact almost none of them keeps is a record of what was decided and why. The reasoning that shaped everything else in the building is the thing that evaporates first.

A decision log is the minimum viable repair: a short, dated record of each significant decision, written at the moment it is made. A decision log is five sentences written at the moment of choice: what we chose, who owns it, what evidence we had, what else we considered, and what we expect to happen. Everything beyond the five sentences is elaboration. Anything less is minutes.

The practice has an honest pedigree. Peter Drucker prescribed a personal version he called feedback analysis: whenever you make a key decision, write down what you expect will happen, then compare the actual result against the expectation months later. Daniel Kahneman has recommended keeping a decision journal as one of the few real defences against hindsight bias, and the investment strategist Michael Mauboussin has done as much as anyone to popularise the habit. Software engineering rediscovered the same idea as the architecture decision record, a format proposed by Michael Nygard: one short file per decision, recording context, choice and consequences. Different fields, one insight. The cheapest way to get better at deciding is to write decisions down while they are still honest.

The five sentences

• The choice. In plain words, one sentence, free of the deck it arrived in. If the choice cannot be said in a sentence, it has not been made yet.
• The owner. One name. Not a committee, not a function. The entry should settle, permanently, the question of who held the decision right.
• The evidence. What was known, and how well it was known: measured, modelled or merely assumed. The honesty of this sentence determines the value of the whole entry.
• The alternatives. What was seriously considered and why it was declined. This sentence protects the future against hindsight’s tidiest lie, which is “we had no choice”.
• The expected outcome, with a date. The sentence most often omitted and the one that matters most, because it makes the decision scoreable. A log without expectations is a diary; a log with them is an instrument.

Why most decision logs die

Most decision logs are dead within a quarter, and the autopsy tends to find the same causes. They are written for audits, not for reuse: when a log exists to satisfy governance, entries are recorded late, phrased defensively and sanitised into inevitability, and a record written to protect its author teaches nothing. The template bloats: five sentences become two pages of mandatory fields, and the cost of writing overtakes the value of reading. The log lives where nobody works, in a register visited only when something has gone wrong. And nothing loops back: without an expected outcome and a date, there is never a reason to return to an entry, so the log becomes write-only memory, and organisations delete write-only memory by neglect.

What makes a log live

The living log inverts each cause of death. It is written at decision time, by the decider, not reconstructed later by a scribe. It is cheap: five sentences, minutes of work, no approval chain. It is revisited on a rhythm: when an entry’s expected-outcome date arrives, someone compares expectation with reality and writes the score next to the original reasoning, which is Drucker’s loop made organisational. Above all, it is read before deciding. The true test of a decision log is not whether people write in it after their decisions but whether they search it before them. That is the moment the log starts repaying its keep, because the next similar decision starts from what the organisation learnt instead of paying the re-derivation tax all over again.

One concrete example

Clearly illustrative, with no customer implied. A firm of a few hundred people declines a large but vaguely scoped engagement. Five sentences enter the log: declined; the managing director owned the call; the evidence was that scope was unbounded in two areas and the client would not phase it; the alternatives were accepting at a premium or phasing the work; the expectation was that the client would return within two quarters with a bounded scope, or the work would prove unprofitable wherever it landed. Two quarters later the entry is scored against what happened. A year later a similar deal arrives, in a different sector, handled by a different team. Someone searches the log, finds the entry and its score, and the new decision begins from evidence rather than instinct. Whether or not the new team decides the same way, they decide knowingly, and that difference is the entire return on a few minutes of writing.

From decision log to Decision Object

A five-sentence log is a practice, and practices are held together by discipline, which erodes. The structured evolution of the decision log is the Decision Object: the same anatomy, held as structured data rather than prose. The evidence sentence becomes facts that carry a quality state from the evidence hierarchy, so what was known is labelled rather than remembered. The alternatives sentence becomes candidate options priced against the constraints that bind them. The expected outcome becomes a score in an Outcomes Ledger, compared against reality without depending on anyone remembering the anniversary. And the collection stops being a diary and becomes institutional memory: queryable at the moment of the next decision, which is what decision intelligence ultimately means by memory.

None of that machinery is a prerequisite. Start with the five sentences, this week, in whatever tool is already open. An organisation that writes its decisions down is already ahead of one that merely makes them.