What can be asked

Reading A Sidebar In One Minute

How to recover both figures from a page in front of you, what to do about the controls that are not facets, and the three mistakes that produce a wrong answer.

What can be askedReference3 to 117 649

Home · What can be asked · Reference

Everything in this record is recoverable from the filter panel of a board, with a pencil, in about a minute. This page is the procedure.

The procedure

One. Count the controls that narrow results by a fixed set of choices. A dropdown, a group of radio buttons, a row of toggles.

To test whether the method improves planning, a small group can visit Monitask directly and include both an ordinary week and an exception. For a separate technical reference, see OpenRefine.

Two. For each one, count its choices. Do not count an "any", "all" or blank entry: that is the option of leaving the filter alone, and the arithmetic already accounts for it by adding one.

Three. Multiply the counts. That is the cells.

Four. Add one to each count and multiply again. That is the questions.

Write the pair down. The board is now described exactly, within this record's bounds.

The three mistakes

Counting the "any" entry as a value. This is the commonest, and it inflates both figures. A dropdown listing "any" plus four real choices is a facet of four, not five. Treating it as five gives one hundred and twenty-five cells where there are sixty-four, and the error compounds across facets.

Counting a sort order as a filter. Sorting rearranges a result; it does not narrow one. A control offering newest, oldest and nearest changes nothing about which listings come back, so it multiplies neither figure.

Counting a free-text box as a facet. A search field is not a facet at all — it has no fixed set of values — and the page on free text in the section on limits is about what it does instead. For this procedure, ignore it.

If a control cannot remove a listing from the result, it is not a facet, whatever it looks like.

The awkward cases

A filter whose options depend on another filter. Some boards show a second dropdown only after the first is set, and its contents change with the choice. That is a different structure from anything here, and the page on going beyond the bounds says so. A rough reading is to take the number of options in the most common case and note that the figure is approximate.

A numeric range with two ends. A minimum and a maximum, each with its own list of choices, are two facets. If they are a single control offering a list of bands, it is one facet with that many bands.

A toggle that means "only show X". Two states: on and off. But "off" usually means "do not filter", not "show things that are not X", so this is a facet of one value, not two — and the page on why a facet starts at two says what that does. It doubles the question count and leaves the resolution alone.

That last case is worth a second reading, because such toggles are common and they look exactly like facets of two.

A worked example

A panel with three controls: a dropdown of six categories, a dropdown of four regions, and a group of six toggles for features, each either set or not.

The first two are facets of six and four. The six toggles are six separate facets — each is its own yes-or-no control, not six values of one facet.

Part	Facets	Sizes
category	one	6
region	one	4
features	six	2 each

That is eight facets, which is past this record's bound of six. The method still works — multiply — but the figures leave the tables. Cells: six times four times two to the sixth. Questions: seven times five times three to the sixth.

The lesson from the awkward case is the one worth keeping: a row of toggles is not one facet, and treating it as one understates both figures badly.

Checking the answer

The pair is self-checking, which is worth using.

Factorise the question count. Each factor of seven is a facet of six; divide them out and the rest should correspond to the remaining facets, each contributing its size plus one. Multiply the implied sizes and compare with the cell count.

If the two agree, the reading was consistent. If they do not, a dropdown was miscounted somewhere — almost always because an "any" entry was counted as a value.

Thirty seconds, and it catches the one mistake that this procedure reliably produces.

What to do with the pair once you have it

Compare it with another board's pair, or with the ranges in this record.

A board resolving under a hundred listings is coarse by the standards of what is possible. One resolving several thousand is middling — the median across the record is eight hundred and sixty-four. One resolving tens of thousands is at the top of the range this record covers.

But the comparison that matters is usually against the other board somebody is considering, and for that the pair is enough on its own. Two pairs, four numbers, two minutes, and the structural comparison is finished and will not need redoing.