How to measure corporate jargon with five document counts and three reader checks

Suppose a team decides to write more plainly, and six months later, has anything changed? Impressions are unreliable, because the people judging are the people who made the effort.

What is needed is a small set of measurements from the documents, plus a second set from readers. This page defines both, explains how to take each, and states what each cannot tell you.

No thresholds appear below. A good sentence length, a good passive share, a good readability score: none of those exist in general, and any figure offered as one is a figure from somebody else's documents for somebody else's readers.

What to take away

Words first

A measure is a count. A metric is a measure standardized for comparison, usually per sentence or per hundred words. A score is a metric turned into one number by a formula, often with a scale.

Everything below is a measure or a metric. None is a score, and none should become a target, for reasons the last section sets out.

An input metric describes the writing. An outcome metric describes the effect on a reader. The first predicts the second, sometimes. Only the second is the point.

Five document metrics

Each can be counted by hand on a sample or by a simple script on the whole

Two of these overlap by design. Actor share and abstraction ratio come from the same pass over the same sentences. Together they tell you the one thing document metrics tell you well: how often the writing says who did what.

That is the grammatical signature of evasive prose described under corporate jargon, and it is worth counting.

Sentence length is deliberately not in the table. It is the easiest thing to count and the least informative, because a long clear sentence and a short empty one are both common. Count it if you like, as a flag for where to look, and never as a result.

Three reader metrics

These are harder to collect and worth more.

Clarifying questions per document. After a document goes out, count the questions that come back which the document should have answered: who, when, what exactly, does this apply to me. Track it over time for the same kind of document. This is the closest thing to a direct measure of whether the writing did its job.

Time to first action. For a document that asks for something, how long until the first reader does it? A request that is easy to find and easy to understand gets acted on sooner. Compare like with like: a request for a signature and a request for a rewrite have different natural timescales.

Misreadings. Cases where a reader acted on a reading the writer did not intend. These are rare, expensive, and the most diagnostic of all. Keep a short log: the sentence, the intended reading, the reading that was taken. Most entries will turn out to be a sentence with no actor, or a word with two senses, and the log will point straight at the document metric that would have caught it.

An illustration, invented

A team samples twenty sentences from each of two memos on the same subject,

one written before the plain-writing effort and one after.

Before: seven of twenty sentences have a person or named body as subject; eleven are passive, of which nine name no agent; four acronyms are undefined.

After: fifteen of twenty have a named actor; three are passive, all with an agent; no acronyms undefined.

That is a real change in the text, but whether it changed anything for readers is a separate question. The team finds out by counting the questions that came back; in the invented case, questions fell from nine to two.

The two remaining were both about a date the memo did not give. The document metrics found the improvement; the reader metric found what was still missing, and neither alone would have.

What readability formulas measure

Several published formulas turn a text into a single score from sentence length and word length or syllable count. They are quick and widely used. What they measure is surface difficulty, computed from features that correlate with reading effort.

They do not read the text. A sentence of short common words that says nothing scores well. A sentence with one necessary long word that says something exact scores worse.

Used as a flag on large volumes, they earn their keep. Used as a target, they produce short, vague sentences. That is the outcome they were meant to prevent.

The federal plain-language guidance on writing for understanding is a useful reference. It treats the passive as a grammatical construction, not a score, which a reliable count needs.

The grammar is set out in the entry on English passive voice. A count that does not follow it will flag every sentence containing "was".

Why none of this should be a target

The rule is general and applies here with unusual force. Set a target for actor share, and writers will insert "the team" as the subject of every sentence.

Set one for passive share, and they will write "we made the decision to" instead of "it was decided", which has an actor and is no clearer. Set a readability target, and sentences get chopped in half at the comma.

Document metrics are for finding where to look: which documents, which sections, which writers might benefit from a conversation. The conversation is the intervention. The reader metrics are the result. The document metrics in between are instruments, and an instrument that is being managed against stops measuring.

The same relationship between a measure, a proxy and a target is set out for business figures generally under business acumen. The words for a metric's five parts, including the denominator that every count above depends on, are under business vocabulary.

Running a check without gaming it

Common questions

Is there a single number that tells me if a document is clear?

No. Clarity is a fact about a reader, and the closest you can get from the text alone is a set of flags for where clarity is likely to fail.

How big a sample do I need?

Enough to see a pattern, which for the metrics above is a few dozen sentences. This is diagnosis, not statistics; you are looking for where to read closely, not estimating a population.

Should we use a readability tool?

As a flag on large volumes, yes. As a score to report or a target to hit, no, for the reasons in the section above.

What about measuring jargon words directly?

A word list catches the obvious cases and misses everything on this page's sibling articles, where the problem is a good word used loosely or a metaphor in place of a claim. Count the grammar, not the vocabulary. The habit of checking whether a message was received, rather than only sent, is covered under communication skills.