A decision is not a frozen file: it evolves
You did the work. The decision was written down, filed in the right place, with its context and its reasoning. Three months later, someone reopens the page and applies what it says. Except the decision changed in the meantime, on a call, in a message thread, or because the customer said the opposite. The page never moved.
Nobody did their job badly. The problem runs deeper: we treat a decision as a file, when a decision is a state that evolves. A file is true or false. A decision is true for a certain period, and then stops being true.
The real cost: a stale decision is worse than a missing one
When information is missing, you know it. You go looking, you ask around, you stay skeptical.
When information is there but stale, you don't stay skeptical. You apply it. That's the trap: a stale decision shows up with all the same signals of confidence as a current one. Same page, same author, same assured tone. Nothing tells you it expired.
That's how a project abandoned for good reasons gets restarted, how a pitch the field already disproved gets used again, or how a new hire diligently applies a rule the team stopped following six months ago. Decision debt doesn't only come from what you failed to write down. It mostly comes from what you wrote down and never revisited.
Why overwriting the page solves nothing
The natural reflex, when a decision changes, is to update the page. Replace the old text with the new one. Clean, current.
Except you just destroyed the most valuable piece of information: the fact that something changed.
The page no longer says you had decided otherwise before, or why you moved, or what triggered the reversal. Six months later someone will propose the original option again, and nobody will be able to say "we tried that, here's what happened". The company has a current page and zero learning.
A decision log that overwrites itself on every edit isn't a memory. It's a photo of the present that erases the past with every shot.
What a real memory does instead: it dates validity
The approach that works comes from memory architectures built for knowledge that changes continuously. The principle is simple, and you don't need to be an engineer to apply it.
You don't delete an outdated fact, you close its validity window.
In practice, every decision carries two pieces of information almost nobody records:
- since when it has been true,
- until when it was true, if something replaced it.
When a new decision contradicts an older one, you don't replace the old one: you close its validity window at the date the new one takes effect. Both then coexist, without ambiguity. One is the current state, the other is the history.
The payoff is double, and this is where it all happens. You get, in one move, what is true today and how you got there. A log that overwrites gives you only the first. A log that piles up without dating gives you neither, clearly.
The three questions a decision must be able to answer
Take any important decision in your company and ask it these three questions. If it can't answer, it's a file, not a memory.
- Is this still true? Is there a date, a signal, an event that would invalidate this decision today?
- What did it replace? What position did you hold before, and what made you move?
- What does it rest on? Does the original evidence still hold, or has the field disproved it since?
The third is the most neglected. A decision rarely becomes wrong on its own: it becomes wrong because what it rested on changed. The customer evolved, the market moved, the assumption no longer holds. If you don't connect a decision to its evidence, you can't see when it just lost its foundation.
What this changes in practice
You don't need a tool to start. You need to change one reflex.
When a decision changes, don't overwrite. Add and date. Write the new decision, note when it applies from, and mark the old one as closed on that same date, with one line on what triggered the change. That line is worth more than all the rest: it's the one that becomes a learning.
And accept an uncomfortable idea: a decision that has never been revisited isn't a solid decision, it's a decision nobody has looked after. The absence of updates isn't a sign of stability, it's often a blind spot.
This follows directly from how to keep track of your decisions: capturing the decision is step one, but capturing isn't enough if what you captured expires in silence. For the full picture, see the organizational memory guide.
The takeaway
A decision isn't a permanent fact. It's a dated state, resting on evidence that can be disproved, and replaceable by a next state.
A business memory worth the name doesn't just store your decisions. It knows when they were true, what they replaced, and it tells you when the ground they stood on just moved.
Anything else is a well-organized folder that lies to you politely.