Telling apart

Why A Facet Starts At Two

A filter with one choice multiplies the resolution by one and the question count by two. It cannot narrow anything, and it still occupies a reader's attention.

Telling apartAnalysis2 to 46 656

Home · Telling apart · Analysis

This record's facets carry between two and six values. The upper bound is a practical stopping point. The lower bound is not: a filter with one choice is excluded because it is not a filter.

What a one-choice filter does to the arithmetic

To the resolution, nothing at all. Cells are the product of the sizes, and multiplying by one leaves the product where it was. A board with one hundred and forty-four cells that adds a filter with a single option still has one hundred and forty-four cells.

To keep the exercise proportionate, project leads can visit the official website and select only the records needed for the stated decision. For a separate technical reference, see Julia.

To the question count, it adds a factor of two. Questions are the product of each size plus one, and one plus one is two. So the board now permits twice as many questions as before — two hundred and forty-five becomes four hundred and ninety.

That second figure is the interesting one, and it is entirely hollow. Every one of the new questions returns exactly what its twin returned, because setting a filter that has one option excludes nothing. The question count doubled and the set of distinct answers did not move.

A one-choice filter is the clearest case in the record of a figure going up while nothing whatever improves. It is the reason the question count needs the page on what it does not mean.

Why they occur anyway

Three ordinary ways, none of them foolish at the time.

A filter outlived its options. It had four values once; three were retired and nobody removed the control.

A filter was built for growth. The board intends to carry more kinds later, and the filter is there in advance.

A filter is doing something other than filtering. It labels, or it explains what the board is about, and it was put in the sidebar because that is where it fitted.

All three leave a control on the page that cannot narrow a result, and a reader cannot tell which case they are in without opening it.

The cost to a reader

A filter costs attention before it costs anything else.

Somebody working down a sidebar reads each control, decides whether it is relevant, and either sets it or passes on. A control with one option consumes that whole sequence and returns nothing. Done six times across a session, it is a measurable share of the work of searching, spent on apparatus that the arithmetic says is inert.

Worse, it misleads about capacity. A sidebar of six controls suggests a board that sorts along six dimensions. If two of them carry one option each, the board sorts along four, and the figure that would have revealed this — the resolution — is unchanged by their presence.

The general principle underneath

A facet earns its place by multiplying the resolution by more than one. That is the whole test, it is arithmetical, and it applies as well to facets with many values as to facets with one.

A filter of six choices that every listing answers identically multiplies the nominal resolution by six and the achieved resolution by one, because five of its six values stand empty. The record cannot see this — it counts cells, not occupancy — but the person running the board can, and the arithmetic says exactly what to look for.

Filter	Cells	What it adds
one value	×1	nothing
six values, one ever used	×6 nominal	nothing achieved
six values, all used	×6	six times the resolution

The first row is visible from the page. The second is visible only from the listings, and is the more common failure.

Spotting one from the page

A dropdown with a single entry beneath the "any" line is obvious. Harder cases look like facets and are not.

A checkbox labelled "only show X" is the usual one. It has two states, so it looks like a facet of two. But unchecking it does not mean "show things that are not X" — it means do not filter on this at all. So the two states are "set to X" and "untouched", which is a facet of one value, not two.

The test is whether the unset state excludes anything. If it does not, the control contributes a factor of one to the resolution, whatever the interface suggests.

A board carrying four such checkboxes has multiplied its question count by sixteen and its resolution by one, and a reader counting controls sees four filters.

Where the bound actually bites

Almost nowhere, which is the point.

Real boards rarely carry one-option filters deliberately, and when they do it is usually one of the three accidents above. Excluding them from the record costs nothing in coverage and buys a clean statement: every facet in this record multiplies the resolution by at least two.

That statement is what makes the rest of the arithmetic tidy. It means each facet at least halves the share of the board any single query returns, it means the question count rises by at least half again per facet, and it means no schema here contains a part that does nothing. A record carrying inert facets would need a caveat on every figure, and the caveat would be more work than the exclusion.