What can be asked

Questions With No Answers

The record counts what a board permits, not what it can satisfy. An empty result and an inexpressible distinction fail differently.

What can be askedAnalysis3 to 117 649

Home · What can be asked · Analysis

A board with six filters of six choices permits one hundred and seventeen thousand six hundred and forty-nine questions. It will not have one hundred and seventeen thousand six hundred and forty-nine useful answers, and probably not a thousand.

The record counts what can be asked. It does not count what comes back. This page is about why that is the right thing to count and where it stops being enough.

Operational context is most useful when managers see the product details and read the output beside the work product rather than on its own. For a separate technical reference, see Microsoft Power BI.

Two ways a search fails

The question returns nothing. The reader asked something the board understood perfectly and there are no listings of that kind. The board has answered, truthfully, that this combination is empty.

The question could not be asked. The reader wanted to separate two things the filters do not distinguish. No query addresses it; the board returns a result containing both and says nothing about the difference.

These feel similar from the reader's chair and they are not similar at all.

An empty result is information. A distinction the board cannot express is an absence of information, and the reader does not know it happened.

Why the first failure is survivable

Because it tells the truth and the reader can act on it.

Somebody who narrows to a combination and gets nothing knows that combination is empty here. They can widen, try a neighbouring combination, or go elsewhere. The board has done its job: it represented the state of its own contents accurately.

Nothing was hidden, nothing was silently merged, and the reader's model of the board is correct afterwards.

Why the second is not

Because the reader's model of the board is wrong afterwards and nothing corrects it.

They set the filters they had, got a list, and worked through it. The distinction they actually cared about was never on offer; the result contains both kinds mixed together, looking exactly like a result that had been properly narrowed.

There is no error message for this. It is the ordinary appearance of a board working correctly, and the ceiling that produced it is the cell count — which the first section of this record is entirely about.

What the question count is good for, then

Comparing boards, not predicting results.

Two schemas permitting two hundred and forty-five and one thousand two hundred and ninety-six questions differ in a fixed, checkable way: the second lets a reader name more collections of the same cells. That comparison holds whatever either board happens to contain, this week or next year.

A figure about how many questions return something would not hold for a week. It depends on the listings, which change constantly, and on the field, which this record cannot see.

Figure	Depends on	Stays true
questions permitted	the schema	until the filters change
questions that return something	the listings	until tomorrow

The record measures the first because it is the one that can be measured exactly and quoted later. The second is a better question and an unanswerable one from the page.

The board's own view of this

Somebody running a board can see which questions return nothing, because the queries arrive in their logs.

That is genuinely useful information and it answers a question this record cannot touch: which combinations people are asking for and not finding. A board that sees repeated empty queries for one combination has learnt something about demand.

What the logs cannot show is the other failure. A reader who wanted a distinction the board does not carry does not issue a query for it — there is no query to issue. They set whatever filters exist, get a mixed result, and leave.

So the logs record one failure in detail and the other not at all, which is why the arithmetical ceiling matters: it is the only instrument that speaks to the silent case.

The honest caveat

A board can inflate its question count with combinations nobody will ever use.

Six filters of six permits a great many questions that are nonsensical in any real field — combinations of values that do not co-occur and never will. They are counted here because the record counts structure, and from the structure they are indistinguishable from the useful ones.

So the question count is an upper bound on expressiveness, and a loose one on a large schema. It is still worth having: an upper bound that can be computed from the page beats an estimate that cannot be computed at all, provided nobody mistakes the one for the other.

Which is the general rule this record runs on. Every figure here is exactly what it says and nothing more, and the places where it stops being enough are marked rather than left for somebody to discover.