Class-Token Coverage is a measurement I developed for [byte-ledger](/glossary/byte-ledger/) audits. The method is straightforward: count the distinct class selectors defined in a stylesheet, then count how many of those classes appear in the rendered HTML of the surface you're auditing, and express the second number as a percentage of the first.
On the thisismyurl.com glossary surface, `main.min.css` shows class-token coverage of 21.1%. That means roughly four out of every five class selectors in the shipped stylesheet have no corresponding element on the page. Every visitor pays the download cost, the parse cost, and the style-recalculation cost for rules that will never match anything they see.
A useful target depends on the surface. For a homepage that introduces every component the site uses, 60-80% is realistic. For a deep article surface that uses a narrower set of components, 30-50% is honest. Below 25% means the stylesheet is doing work for surfaces it isn't on, and the fix is per-surface enqueuing or a smarter build that ships only the rules each template needs.
What class-token coverage doesn't measure: rules that match elements but have no visible effect, rules overridden later in the cascade, rules inside `@media` queries that won't fire on this viewport. It's a coverage signal, not a completeness signal. Pair it with the rest of the byte-ledger for the full picture — coverage tells you what to investigate, the ledger tells you what to refuse.