The Data Park · a four-minute walk
When is it data?
A walk of about four minutes, starting from a single number and ending at the smallest serious thing you can say about it. You will be asked to commit to an answer before each reveal. That is the whole method; guessing wrong is the fastest way through.
This is a teaching walk, not a publication. It simplifies on purpose, in places a serious reader will want undone — and the account it simplifies is at the far end, in its own clothes.
The specimen
Before you scroll — what is it?
one binding later
age = 42
Something arrived. Be precise about what.
A kind is the cheapest thing to add and the first thing people mistake for meaning. It rules out nonsense. It does not yet say what the number is about.
two bindings later
John's age = 42
Now it is about somebody. Is it data?
Which is why the honest answer here is it depends on what is around it — and why a number alone on a plinth, with nothing around it at all, is the one case where the answer is always no.
the demonstration · operate the lever
Five stored values. One column. Add them up.
Every value below is stored. Every value is typed as a number. Your database will add this column without hesitating and without warning you. Go on — the total is right there.
Nothing here was corrupted. That is the part worth sitting with.
Every digit survived. Every type survived. The column is exactly as sound as it looks, and it would still be sound in ten years. What was never written down is the one thing that would have told you these are five different quantities that happen to be spelled the same — and so the machine did the only thing it could do with five numbers, which was add them.
That is the whole failure, and it is worth saying flatly: representation was kept perfectly and identity was lost entirely. Nothing warns you, because from where the software is standing, nothing went wrong.
fold out About row five
Deep Thought computes for seven and a half million years and returns 42. The joke is usually filed under absurdism. It is not. It is a joke about data.
The machine's answer is correct. Nobody can use it, because nobody knows the question — and the question is precisely the thing that would have said what the 42 was about. Adams built a seven-and-a-half-million-year computation that produces a perfectly typed value at no analytical point whatsoever, and then made everyone in the scene react exactly the way an analyst reacts to a number in a spreadsheet with no column header.
The sequel to that joke is that they then have to build a second, much larger computer to recover the question. Which is roughly what a data-governance project is.
the smallest serious formulation
A datum is a typed value at an analytical point.
- typed
-
what kind of thing the value is — and therefore what may lawfully be done to it. This is
the part
age = 42supplied. - at an analytical point
- which thing, in which world, at which grain. It does not have to be written on the value; it has to be determined by the world the value sits in. This is the part five identical 42s did not have, and the part no storage layer supplies for you.
A number is not data until its point is determined — until what it is about, and where in the world that is, are fixed by something other than the reader's assumption.
sealed · an open question, and we mean open Turn it around. If the analytical point is perfectly determined and no value was ever observed — what exactly is missing? break the seal
Everything up to here has been a value looking for a point. This is a point looking for a value, and it is much harder.
A shop that was open on Tuesday and sold nothing. A shop that was open and whose till failed to report. A shop that was shut for refurbishment. A shop that does not exist. Four situations, four different answers to “what is missing”, and a familiar record-model instinct flattens all four into the same thing: no row. Or, if someone was being helpful, into a zero — which is an answer, confidently given, to a question nobody checked was askable.
The separation itself is settled. “Missing” is at least three separable things — the point does not exist, the point exists and is not eligible, the point exists, is eligible, and was not observed — and that is doctrine, not an open question. The open part arrives the moment you separate them, because somebody then has to say who is entitled to decide which one you are looking at, and what an honest answer should look like in each case.
That frontier — authority over the distinction, and what an honest response owes you in each case — is where our own work sits, and it is not finished. It is also the reason a declared world can contain an address that lawfully exists with nothing behind it: a sentence that sounds like a bug until the day it saves you.
It is also, if you want the honest version, the next thing we would build a walk about.
↳ the deeper trail — where “analytical point” goes if you follow it
Three sentences from a longer theory. This is a gesture at the shape, not the account.
universe
A set of points, plus the law that says when a point exists. Orders exist by occurrence — an order is a point because it happened. Store-days exist by declaration — the pair (store 12, yesterday) is a legitimate location whether or not anything happened there. Two different kinds of existence, and confusing them is how a denominator drifts.
anchor
A governed partition of a universe, written as its coordinates:
{store, day}, {store, month}.
Coarser and finer anchors are the same territory re-tiled, which is why some rollups are
free and others are not lawful at all.
the complete name
A quantity with its law, at an anchor: revenue @ {store, month}.
That is the whole shape this walk has been climbing toward — the family, at the address.
Our bare specimen wanted to be
age @ {person, date}, and everything that was missing from it
was missing from exactly one of those two halves.
If that landed, the next thing to read is not another attraction. It is at the bottom of this page.
you are now leaving the park
The actual account
Everything above is a teaching device. It is deliberately smaller than the theory, and it simplifies in places a serious reader will want undone. These do not.
- The Theory of Data The foundation itself, with the proofs.
- A Primer on the Theory of Data The vocabulary, carefully, in dependency order.
- The Theory of Data in One Afternoon The whole idea in one sitting, from a cold start. A teaching artifact hosted here, not a deposited publication.
The park exists because some ideas are easier to walk into than to read into. It does not replace the corpus and is not evidence of anything.