Why a level costs more than a sibling
A category tree has two ways to grow and they are not comparable. Adding a sibling at the top level adds one row to one page: the reader scans a slightly longer list. Adding a level multiplies — every node that already exists at the bottom gains a set of children, and each of those children needs a name that is not a synonym of its siblings, a slug nobody else holds, a sentence saying what belongs in it and a sentence saying what does not.
- 1Twelve categoriesOne page, twelve rows. A thirteenth is one more row.one click
- 2Seventy-two nodesSix children each. Every one needs a name and a slug.two clicks
- 3Four hundred and thirty-two moreThe same shape again, and the naming bill with it.three clicks
The planner on the home page prints that multiplication because the number is the argument. Nobody agrees to four hundred and thirty-two naming decisions in the abstract, which is why the arithmetic is shown before the level is agreed rather than after.
The three-click budget, and the second route
Every category sheet carries the same cap: the deepest leaf is three clicks from the home page. Three is not a rule about reading speed, it is a rule about abandonment — past three steps a reader stops believing the thing they want is down there at all.
When a tree genuinely needs a fourth level, the budget is not waived; a second route is specified instead, and the sheet has to name which of the three it is.
One parent only. A category has exactly one parent. A grouping that cuts across the tree is a facet and lives on the search surface, because a node with two parents has two addresses, and two addresses mean the same page competing with itself.
The three second routes
A facet on the search surface that reaches the leaf without walking the tree.
A cross-link from the sibling the leaf is most often confused with.
A search row that lands on the leaf directly, from a term the leaf’s own sheet lists.
What a category sheet carries
- 01 · nameIn the words a reader would use rather than the words the trade uses.
- 02 · slugThe address, and the date it entered the register.
- 03 · parentThe single parent, by identifier.
- 04 · inclusion testOne sentence that is true of everything filed here.
- 05 · exclusion testOne sentence naming what gets filed here by mistake, and where it goes instead.
- 06 · confusable siblingNamed, with the cross-link that resolves it.
- 07 · second routeOnly if the node sits past the budget.
A rename changes the name and leaves the slug alone, because the slug is an address and the name is a label. A merge is the heavier move: children are reparented one at a time, both slugs stay in the register with the loser redirecting to the winner, and the record keeps the row that says which way round it went. Nothing in the tree is deleted outright — a deleted register entry is a broken address with no way back.
Words the surfaces answer on are reserved and cannot be taken by a category. The checker on the home page flags them alongside the collisions, since both failures end the same way: a reader types an address and does not arrive at the category.