The visibility temperature scale

A visibility temperature runs from 0 to 100 degrees and falls into one of five bands: Freezing, Cold, Warming Up, Warm, Hot. The band is the part you remember; the number is the part you compare.

The bands exist because a number on its own invites the wrong question. Told they scored 62, people ask how to get to 63. Told they are Warming Up, they ask what warm looks like. The second question is the useful one, and the band is what prompts it.

• 85–100°

  Hot

  Everything observable is in place.

  Every signal a thermometer can see from outside is present. From here the work stops being technical and starts being editorial: what the page actually says, who else cites it, and whether it deserves the position it is asking for.

• 65–84°

  Warm

  Working. Not yet hard to beat.

  The page states what it is, who published it and what it answers, and a retrieval system can use it without guessing much. What is left is usually a handful of deliberate signals rather than anything broken.

• 45–64°

  Warming Up

  The basics are in, the signals are thin.

  Most of the ordinary things are present and several of the ones that separate a page from its competition are not. This is the most common reading on the internet and the one where a small number of specific changes moves the number the most.

• 20–44°

  Cold

  Readable, but nothing is helping it.

  The page exists and can be fetched, and that is roughly where the help stops. Machines are reconstructing what it means from prose alone, which is where misrepresentation starts and where competitors with the same content but better signals go past you.

• 0–19°

  Freezing

  Almost nothing is getting through.

  At this temperature a search engine or an answer engine is working with raw text and little else. Something structural is usually wrong: the page is blocked, it is built entirely in the browser, or it never returned to the crawler at all.

Why the bands are not the same size

Cold runs 25 degrees wide and Warming Up runs 20, because the middle of this scale is where the interesting distinctions live and the bottom is where they stop mattering. A site at 8 degrees and a site at 18 degrees have the same problem, which is that something structural is broken. A site at 48 and a site at 63 have genuinely different problems, and a band wide enough to hold both of them would be useless.

The top band starts at 85 rather than 90 for the opposite reason. Past 85 every signal a thermometer can see from outside is either present or so minor that arguing about it is a way of avoiding the harder work. There is no meaningful difference between 92 and 97, and a scale that implied there was would be selling you something.

What a temperature is not

It is not a prediction. Nothing here forecasts a ranking, a click, or a sale. Two pages at 68 degrees can perform completely differently, because the thing separating them is competition, demand and whether the writing is worth reading, and none of those are visible from a fetch.

It is not a grade either. A brochure page for a local business at 55 degrees may be doing everything it needs to. A publisher competing nationally at 55 is in trouble. The number describes the page, and you supply the context.

What it is good for: deciding whether to look closer, comparing two pages on the same basis, and watching one page change over time. Those are the three things a thermometer is for.

Taking the same reading twice

A reading is reproducible in the sense that the same page in the same state returns the same temperature. It is not frozen: the page can change, the site's robots.txt can change, the server can be slow on the day. A reading that moves without anyone touching the page is usually the response time signal, and that is worth knowing too.

Every published reading in the readings dataset carries the date it was taken for exactly this reason. Re-take any of them yourself and compare.