Dimension 03 · Clarity
Would someone actually understand what they are agreeing to?
Give this a privacy policy and it measures how hard it is to read, how much of it there is, and what that adds up to on the Data Human Index. Paste it, open a PDF, or hand it a web address.
What is in the document
How hard it is to read
The Clarity score
The thresholds
Where the numbers turn into a score
These cut-offs are judgments rather than arithmetic, so they are published here instead of being buried in the code. They are calibrated against the policies actually scored on the Index, which run from roughly 4,500 to 8,700 words at grades 10.4 to 13.2.
The first version of this page aimed at plain-language ideals instead: Grade 8, 1,500 words. Every real privacy policy fell into the bottom two boxes, every brand scored the same, and a dimension where nothing differs measures nothing. These thresholds separate documents that genuinely differ. Revised 6 August 2026.
| Consensus grade | |
|---|---|
| 4 | 9.5 or below |
| 3 | 9.5 to 11 |
| 2 | 11 to 12.5 |
| 1 | 12.5 to 14 |
| 0 | Above 14 |
| Words | |
|---|---|
| 4 | Under 3,000 |
| 3 | 3,000 to 5,000 |
| 2 | 5,000 to 7,000 |
| 1 | 7,000 to 9,000 |
| 0 | Over 9,000 |
Clarity is the mean of the three components, rounded. Length is included because difficulty alone rewards the wrong thing: a policy can be written in short, simple sentences and still be unreadable because there are nine thousand of them.
What this does not do
The limits, stated plainly
Readability formulas count syllables, words and sentences. They cannot tell whether a sentence is honest, whether an important term is defined somewhere else, or whether the difficult part has been moved to a linked document. A policy can score well here and still mislead, and Clarity is one dimension of five for exactly that reason.
The formulas also disagree with each other, routinely by several grades, which is why the median is used rather than any single one. Treat the consensus as a rough position rather than a precise reading.
How text is broken into sentences changes the answer more than most people expect. A PDF wraps prose at the page margin, and a tool that treats each wrapped line as its own sentence will report a document as far easier or far harder than it is. This one rejoins wrapped lines and starts a new block only at a bullet or a numbered item. The same logic runs in the command line version used to score brands, and the two are checked against each other, so the page and the script cannot drift apart.