A List You Cannot Recite

How many things a company can hold constant, and where the rest belong

A company that decides to be deliberate about what it protects tends to produce a long list. Every entry on it is defensible, everybody who contributed feels heard, and most of the entries will never be defended by anybody. The length is why.

Nothing below is measured, and it comes from reading scaling companies closely. The research cited later supports the argument and is not the argument, and the piece says where the line falls.

Defence happens in the room, from memory, while everybody is in a hurry

The moment a commitment is tested is not a review meeting with the document open. It is a Thursday, in a conversation about something else, when somebody proposes a change and one person has to notice that this is the thing.

Nothing is open in front of them. The register is in a wiki nobody has visited since the offsite, and the conversation will have moved on by the time anyone could find it. What the person has is whatever is already in their head.

So the operative constraint on a kept list has nothing to do with tidiness or with executive attention span. It is that defence is performed from memory, under time pressure, by somebody who was thinking about something else a moment earlier. A list that cannot be recited under those conditions has no mechanism.

There is relevant research and it is worth stating carefully. Nelson Cowan's 2001 synthesis in Behavioral and Brain Sciences, reviewing evidence across verbal, visual and spatial tasks, put working-memory capacity at around four chunks, revising the older figure downward. That is a finding about a laboratory task rather than about company policy, and it is offered here as a reason the operational limit sits where it does rather than as proof of it. The claim CloudFruition can defend is the one about the Thursday.

The completeness objection is the right instinct pointed at the wrong artefact

The strongest objection to a short list is governance-minded and it is correct on its own terms. An unlisted commitment is an unmanaged one, and a company that genuinely relies on a dozen arrangements has an incomplete register the moment it records three of them, whatever the presentation.

That objection is right that the remainder matters and wrong that this list is where it goes. It treats one artefact as though it had to do two jobs: hold everything the company knows, and be present at the moment of defence. Those requirements pull in opposite directions, because a document complete enough to be authoritative is too long to be carried in somebody's head.

Most functions solved this long ago by splitting it. There is a register, and there is the short thing people actually remember. The mistake is running both jobs through one page.

Where everything that comes off the list goes, which is what makes the cap honest

Nothing gets discarded. Everything that comes off routes somewhere, and knowing where is what makes the cap defensible rather than arbitrary.

Some of them are decisions rather than commitments. They have not yet become things the company relies on, so they belong in the ordinary sort by how hard they would be to undo, which allocates scrutiny at the point of deciding and is a better fit for them.

Some of them are already protected by their own cost. An arrangement that would take a year and a migration to unwind does not need a place on a list somebody recites, because nobody removes it on a Thursday. It is defended by its own weight.

Some belong in a written record rather than a memorised one. A decision record, holding what was decided, the evidence, who owns it and when it gets revisited, is the right home for an arrangement that needs documenting rather than defending in the moment.

And some, on inspection, were preferences. They survive the exercise as things the company would like, which is a real category and a different one, and no harm is done by saying so.

The short list is what remains: arrangements that are genuinely relied on, cheap enough to remove that somebody might, and important enough that removal should cost a conversation. That set is small in most companies, and the smallness is a property of the world rather than a design choice.

What to do this week

Write down everything your company means to hold constant and let the list run long. Ten minutes, and do not edit while you write.

Then take each entry through the four routes above and see what is left. Decisions to the sort. Expensive ones to their own weight. Documentable ones to a record. Preferences to the honest pile.

Whatever survives, close the exercise the same way: put the list away and ask two people to say it from memory a week later. What they can say is the list. What they cannot say is a document, and a document is not in the room on the Thursday.

Which of the things your company protects could the person who has to defend it name without looking?