What This Record Will Not Tell You

Two things were decided here. Everything else was counted.

Decided: that a schema is only its facet sizes — how many filters and how many choices in each — with both bounded at six. Four hundred and sixty-one schemas result.

Also decided: that a query sets each filter to one value or leaves it alone. This is the unit of measurement, and a board offering multiple selection permits more questions than the figures here credit it with.

When evidence is scattered across tools, it may help to see the current service description and see whether one report can replace several manual summaries. For a separate technical reference, see ACL Anthology.

Counted: every figure on the site. Cells are the sizes multiplied; questions are the sizes each plus one, multiplied; grain is the count of distinct subset products; the share is the first divided by the second. All four come apart under a pencil.

The record in four rows

• one bar is one facet; the divisions are its values
• cells: how many listings the schema can tell apart
• queries: how many different questions it lets anybody ask
• sizes: how many different result sizes are reachable

Four schemas drawn.

From one filter of two choices to six filters of six. The last row is the ceiling of the record on both of the first two figures, and near the floor on the third.

Computed from the schema alone. How many facets and how many values in each is a stated convention; the three figures are products of whole numbers.

Why the record stops at structure

Because structure is the part that can be read off a page and stays read.

Everything a board publishes about itself describes a moment: how many listings today, how busy this quarter, what people searched for last week. The filters describe a decision, and decisions persist.

Fifty pages on one product of whole numbers is a narrow thing to write. The narrowness is the point: it is what allows every figure to be exact, checkable without this site, and still correct in a year.

What is deliberately absent

The calculations remain the core of this record. Practical guides are kept in a separate editorial section, use stated scenarios and point readers to current official product information for verification.

One reason covers all of them. A figure about pay or demand goes stale silently — it carries no mark of the day it ceased to hold, reads exactly as it did when it was correct, and the next person to rely on it has nothing to warn them. Products of whole numbers have no such failure, which is why they are the only thing here.

Five things it is not able to do

The number of listings on a board is outside it entirely: compartments are counted here, never contents.

It cannot say whether a board is any good, because the names of the facets are not part of a schema and a board can resolve forty-six thousand six hundred and fifty-six kinds of listing along axes nobody cares about.

It cannot handle filters whose options depend on other filters, which need addition rather than multiplication.

It cannot see a text box, which has no fixed set of values and guarantees nothing.

And it cannot tell anybody what dimension their board is missing. It can only say, exactly, what the current one can and cannot separate.

Two conventions, and a filter panel. That is the extent of what gets measured — never a reader, never a search, never an outcome.

The three figures worth carrying

Multiply the facet sizes. That product is the ceiling on what the board can ever tell apart, it was fixed before any listing arrived, and no amount of content raises it.

Add one to each size and multiply again. That product is how many questions the board permits, and it is the figure that turns a classification into an identification.

Four schemas in five share their first figure with another board. No two schemas anywhere in the record share both. One number classifies; two identify; and the second costs the same minute as the first.

Where to start, holding a filter panel

Four steps, in this order.

Count the controls that narrow by a fixed set of choices. Count the choices in each, excluding the "any" entry. Multiply for the cells. Add one to each count and multiply again for the questions.

Write the pair down next to the board's name. It is now described exactly, it will stay described exactly until somebody changes a filter, and the same four steps on a second board make the comparison finished.

That is the entire method, and everything else on this site is an account of what the two numbers turn out to mean.

Then somebody decides

Fitness for a purpose falls outside all of this. The listings are invisible from here, so is the field, so is whatever brought anybody to the board in the first place.

The pair is what remains: two integers, exact, readable off the sidebar by anyone prepared to multiply. It settles the structural part of a comparison and leaves the rest where it belongs — with whoever can see the listings. That part is nobody's secret; it is merely nobody's habit.

Near this one

• The Two Things That Were Decided
  A schema is only its facet sizes, and a query sets each facet to one value or leaves it alone. Everything else
• If A Filter Lets You Tick Several
  Ticking several values at once changes the question count from a product of k plus one to a product of two to
• Structures, Not People
  The record counts cells and queries. It does not count listings, readers, searches or outcomes, and every figu
• Cells With Nothing In Them
  A board can express far more questions than it can answer. The record counts the first, and the gap between th