What things mean
Our other index tells you where a file lives. Atlas tells you what a thing IS and how it relates to everything else — the companies, the content, the campaigns, and the labels that tie them together.
One clear, verified description of the things we track and how they connect — written to be read by a person or a machine.
BrandTrackers is a library of brand and marketing work — companies, the content they make, the campaigns they run, and the labels that describe all of it. Over time, the logic of how these pieces fit together has lived in scattered documents and in people's heads, and it kept getting re-explained.
Atlas is the one place that logic lives. It describes each part of the domain in plain words, says how the parts connect, and points to the exact data behind each claim. Read it before making a big decision about how the product is structured.
Everything we track falls into one of four groups, and this is how they hang together:
Click a box to read that pillar. Three of the four families are written; the last follows the same shape.
Each pillar follows the same pattern you'll see in Entities: what it is, in plain language; how it connects; a real worked example generated from live rows; and where to look for everything else.
A page here answers exactly one kind of question. A document that mixes them makes you decide, sentence by sentence, whether you are reading a rule, a measurement, a defect, or a musing — and that is a tax every reader pays forever.
| Surface | Answers | Changes when |
|---|---|---|
| The model | What is this thing, and how does it connect? | a decision is ratified |
| Decisions | What is settled, and what is still open? | a question is ruled on |
| Reference (generated) | How much is filled in, and where does reality disagree with the model? | the data or the model moves |
| Internal notes | How is this being built? | any working day |
One rule follows from it: a model page contains no unresolved question and no bare number. If something is undecided it lives in open questions. If it is a count it is generated into the state of the data. If it is a place where the system contradicts the model, it is a check on the conformance report — not a paragraph that quietly goes stale and that nobody remembers to delete once it is fixed.
That is what makes a model page reviewable. It can only be wrong about meaning, which is the one thing reading it can actually catch.