On closure as a discipline.

An institution is a set of stories.

Every organisation, public or private, is in fact a great quantity of small stories running side by side. A vendor invoice arrives. A patient is admitted. A permit is requested. A complaint is lodged. Each one opens a thread of work, asks a number of actors to do their part, and waits — with varying patience — to be closed.

Read an institution that way and a strange thing comes into focus: most of the stories never close cleanly. They are marked finished while the work is still moving. They are routed to a queue and quietly forgotten. They are closed on paper, but the proof is not in the ledger. The institution does not know, precisely, which of its stories are open. It does not know, precisely, what stalls them. It cannot quite say what it has and has not done.

This is not, in our experience, a failure of effort. The people inside the institution are working hard. The systems are running. The dashboards are green. The failure is a quieter one — the institution lacks a discipline of closure. It begins many more stories than it ends, and it has no reliable way to read the difference.

Why guessing is not enough.

The dominant answer the industry offers to this kind of trouble is to fit a model. A great quantity of data is fed to an algorithm. A loss is minimised. A confidence is reported. The institution buys the result and is told: this is what will happen, with this probability, if you act so.

We have nothing against models. They have their place. But a model is, in the end, a guess about what will happen — sometimes a very fine guess, made on solid ground. It is rarely an answer to the question the institution is actually living: what is the case in here, right now, in the work we are doing?

A model says what will happen. An observation says what is the case. The institution needs both — but it has, almost everywhere, only the first. — On the discipline

The first cost of guessing without observing is that the institution learns slowly. When a prediction fails, no one is sure quite why; the model is retrained; the loop closes nothing. The second cost is that nobody owns the failure. The model misfired. The data drifted. The story that was open is still open. No one is required to leave proof.

What an observatory does.

We are starting, at Sobus Hub on Freston Road, an observatory. The work it does is unfashionable. It does not begin with a model. It begins with a patient act of looking — at the institution as it is — and a slow accounting of every story it opens.

Three movements run through the work. The first is to observe: to sit with the people doing the work, to read the systems they touch, to walk the ledgers and the tickets and the hand-offs the institution already records but rarely studies. The second is to diagnose: to name, plainly, where the stories fail to close — which loops are waiting on a proof that never comes, which acts are taken without record, which closures are declared on paper while the work is still moving. The third is to close: to design, for each kind of story, the small protocol that finishes it — the proof its actor must leave, the rule that returns it to observation, the instrument the institution can keep running once we have gone.

We leave artefacts at each step, all of them readable by a person who is not a specialist. A specimen, after the first movement: every story the institution opens, named, with its real path traced. A reading, after the second: the operational holes, ranked by what the institution loses each time a story stays open. A protocol, after the third: one page per story-type, plus the instrument to run it. Nothing in any of these documents requires the institution to take our word for anything. The proof is in the ledger.

The conservation law.

A discipline needs a small law it can be held to. Ours is one line.

k + i = 4.

It is the conservation law every observation we make must satisfy. k counts the structure we have unfolded out of the data — the part the institution can now see. i counts what remains hidden — the part the institution does not yet have a reading for. The two together are fixed. We cannot inflate the visible without admitting more is hidden; we cannot claim less is hidden without showing more visible structure.

The point of the law is not its arithmetic. The point is that two independent viewpoints — the algebraic and the geometric — must agree, point-for-point, or the observation is rejected. There is no model knob to turn. There is no confidence to negotiate. Either closure holds, or the work returns to observation and is done again.

This is what we mean by a discipline of closure. A small law the work can be held to. A proof the institution can read. An act that returns the story to the ledger as truly finished — or as a smaller, named open the next person can pick up.

What we refuse to do.

A founding note is also a place to be plain about refusals, since they shape the work as much as the promises.

1. We will not predict. We will name what is the case. If the institution needs a forecast on top of that, it can build one — on ground that has been observed first.
2. We will not score. There is no confidence number on our work. An observation is either closed by its own proof, or it is rejected and done again. The institution receives a yes or a no.
3. We will not stay. The instrument we leave is one the institution can run on its own. We will not become a permanent dependency. The point of a closure is that someone else can carry it.
4. We will not flatter. A reading that tells the institution only what it wants to hear is not a reading. We are paid for the holes we find, not the comfort we give.

The promise we leave.

The promise of an observatory is, in the end, a small one. We will arrive at an institution. We will sit with the work. We will name the stories it opens and the ones it has not closed. We will leave behind a protocol — one page per story-type — and an instrument the institution can run on its own. We will not be in a hurry. We will not pretend.

We are starting in London because London is the institution we know best — a city that is, more than most, a great quantity of small stories running side by side. Every borough is a chapter. Every procurement loop, every patient hand-off, every permit, every complaint is a story the city has opened and may or may not have closed. Volume I is the city. The folios that follow will be the first studies it allows us to publish.

If you are the steward of an institution and any of this resounds — if you suspect the dashboards are greener than the work — you may write to us at the address on this page. We will read the letter on a Tuesday. We will reply within the week, unless a Volume is being bound.

Quietly. Repeatedly. With evidence.