Nothing to check

A Share Is Not a Risk

The blind share counts possibilities, not likelihoods. Reading it as a probability needs an assumption the record declares and does not make.

Home · Nothing to check · Caution

Fifteen blind contents out of thirty-six is five in twelve. Fifteen out of one hundred and twenty-six is five in forty-two. One out of eighty-four is one in eighty-four.

Three shapes on the same page, three shares, and not one of them is the chance that anything happens to anybody. If the same issue recurs across projects, managers may review the current product page and compare the resulting evidence with finished work.

Same blind count, three different shares

the whole bar: every content a reader might come away with
the dashed part: contents holding nothing checkable

Three shapes drawn.

The first two rows have fifteen blind contents each, over contents figures of thirty-six and one hundred and twenty-six. The third has one, over eighty-four.

Computed from the three numbers. A blind content is one holding no checkable statement: that is a property of the page, not a prediction about any reader.

What the share actually counts

It counts contents. Of all the r-sized subsets a reader might come away with, it is the proportion that contain no checkable statement. For additional background relevant to the interpretation of online work and information, see OECD.

Turning that into a probability requires one further step: assuming every subset is equally likely to be the one a reader gets. The record declares that assumption as the second of its three conventions and does not defend it, because it is not true.

Readers do not sample uniformly. They meet what is first, what is prominent, what a search result surfaced, what a colleague sent them. The real distribution over contents is heavily weighted and nothing in this record has access to it.

Which direction the error runs

It depends on where the verifiable statements sit, and the record cannot tell.

If the checkable statements are prominent — early, headed, in the places readers land — then the true chance of a blind encounter is lower than the share, perhaps far lower. The share overstates the problem.

If they are buried — a line in a table, a figure at the end, a qualification in small type — the true chance is higher than the share. The share understates the problem, and the shapes with the most reassuring fractions are the ones where the understatement is proportionally largest.

The blind share is a count of ways. A risk is a count of ways weighted by how often each one happens. The second number needs information this record does not have and will not invent.

Why the record stops short of a probability

It would be easy to assign plausible weights and publish an expected blind share per shape. The result would look more useful and be worth less.

Every figure downstream would then carry two components — the arithmetic and the weighting — in a single number, with no way for a reader to separate them. A reader who disagreed with the weights could not recompute without knowing them, and if the weights were published alongside, the result would simply be this record plus an extra table of guesses.

What the record does instead is state the convention once, apply it consistently, and print the inputs beside every output. A reader who has their own view about prominence can take the counts and weight them themselves. The counts are the part that can be got right.

The one figure that does not need the convention

The threshold result, again, because it is the only thing here that survives.

When fewer statements are unchecked than a reader meets, no blind encounter exists. No weighting can create one, no assumption about prominence matters, and the conclusion holds for every reader on every path. It is the third row's neighbour on this page: nine published, six met, and with four or more checkable statements the blind count is nought outright.

Below the threshold the record offers counts and a declared convention. At or above it, the record offers a fact. The difference between those two should be visible in how any sentence here is phrased, and where it is not, the phrasing is at fault.

The same objection applied to the opening count

Everything above holds for the count of checkable openings and holds harder.

A blind share assumes contents are equally reachable. An opening share assumes the first statement a reader meets is a uniform draw from everything published, which is a stronger assumption about a narrower thing: entry points are the least random part of any encounter, being decided by links, search fragments and whatever is quotable.

So of the record's two verifiability figures, the one that looks simplest — the checkable count over the published count — is the one leaning hardest on the convention. The next part counts it anyway, states the objection three times, and keeps the figure because a floor with a known weakness is worth more than silence.

The sentence this note exists to prevent

"In one hundred and ninety-nine of three hundred and thirty-eight shapes, a reader will come away unable to check anything."

Will. That word converts a count of possibilities into a prediction about people, and nothing in this record licenses it. The defensible version is: in one hundred and ninety-nine of the three hundred and thirty-eight shapes tabulated here, such an encounter is possible, and whether it occurs depends on which statements readers actually reach.

Longer, duller, and the only version that survives being checked.