Numbers in Running Prose

B · Plate 07 in the Plain Language series · Other Communications

A number in a sentence does one of two things. It lets the reader judge a magnitude, or it borrows the appearance of precision for a claim that is not really quantitative. The difference is usually visible in whether the number has anything to compare itself to.

An isolated figure is nearly meaningless to a reader who does not already work in the field it came from. A rise of four hundred is large or small depending on what it rose from, how fast it usually moves and how accurately it is measured. Supplying the figure without those three is not informing; it is decorating an assertion with a digit.

The comparison the reader needs

The most useful companion to a number is the previous value of the same number, stated in the same units in the same sentence. The second most useful is the total it forms part of. Percentages are the least useful of the common forms when given alone, because a percentage conceals both its base and its absolute size, and a reader cannot reconstruct either.

Where a percentage is genuinely the right form, giving the base beside it costs five words and removes the ambiguity entirely. Where both a percentage and an absolute change are available, giving both is not redundant: they answer different questions, and different readers arrive with different ones.

Rounding is an editorial act

Rounding is not neutral. It signals how much confidence the writer has in the measurement, and readers pick up that signal whether or not it was intended. A figure given to two decimal places asserts a precision that most real-world measurements do not have, and when the underlying method is approximate, the decimals are a small untruth.

The reverse error is rounding away a distinction that matters. A quantity reported as about a thousand when the difference between nine hundred and eleven hundred changes the conclusion has been rounded past the point of usefulness. The right precision is the precision at which the reader's decision would change, which requires knowing what the reader is deciding.

Where numbers should not be

Some numbers do not belong in running prose at all. Any sequence of more than three figures compared across more than two categories is a table pretending to be a sentence, and the reader will have to build the table mentally in order to follow it. Building it for them in the text is not a design compromise; it is the honest form for the content.

Similarly, a number whose only function is emphasis should usually be removed rather than softened. A claim that something happens constantly is at least clear about being an impression. The same claim dressed as nine times out of ten, with no measurement behind it, is an impression wearing the clothes of evidence, and a reader who checks will not check gently.

Everything in this essay reduces to one question asked of each figure in a draft: what could a reader conclude from this that they could not conclude without it. Numbers that survive the question earn their place and usually need a companion figure to do their job properly. Numbers that do not survive it were serving the writer's confidence rather than the reader's understanding.