Framework
Performance, Experience, Emissions and Ranking. The framework behind every audit we run and every site we ship, including this one.
One number would hide the thing you needed to know.
A single score is easy to put on a slide and useless to act on. A website can be quick and unusable, or beautiful and invisible, or all three of those and still cost more to serve than it needs to. Averaging them produces a number that moves for reasons you cannot see.
So P.E.E.R. keeps them apart. Four scores, each measured its own way, each with its own fix list. They are related — a lighter page is usually faster, and a page a screen reader can follow is usually one a crawler can too — but they fail independently, and the work to repair each one is different.
What each one actually measures.
01
What the page costs a real visitor on a real connection: time to something readable, time until it stops moving, and how much has to be downloaded and executed before either happens. Measured against Core Web Vitals rather than a lab score taken on a fast machine.
02
Whether the thing can be used. WCAG 2.2 conformance, keyboard operation, whether it works on a phone held in one hand, and whether the important action on a page can be found without hunting. An automated pass with axe-core, then the criteria a machine cannot judge, checked by hand.
03
Energy measured rather than estimated: real browser CPU-time, converted using live grid data, reported with a confidence range instead of a single tidy figure. Weight is the lever, because the saving repeats on every visit.
04
Whether a search engine or an assistant can read the page, understand what it claims and attribute it. Structure, markup and the machine-readable layer, not keyword density.
The part that makes it worth having.
Anyone can score a website out of ten. The question is what the number rests on, and whether it survives somebody checking it.
The emissions half is the clearest example. It would be easy to publish a single gram figure and a global per-visit average alongside it, and a great many sites do. We do not, because an average tells you about the web rather than about you, and a bare number with no range implies a precision the method does not have. What comes out instead is a calibrated estimate with a 90% confidence range and a stated honesty tier, so you can see how much of it is measurement and how much is model. The method is published.
The performance and experience halves are the same discipline pointed at different evidence: published research on what speed is worth, in the science behind the framework, and a standard with testable criteria rather than an opinion about usability.
We run it against this site before we run it against yours, and publish the result whether or not it flatters us.
Three places, same four scores.
01
A scored audit of an existing site, with a fix list in the order worth doing it rather than the order we found things.
02
The same four are the acceptance criteria for anything we ship, which is why they are not a phase we quote for separately.
03
Scored on the same framework and published in the footer, because a standard you exempt yourself from is not a standard.
Get started
Send us the address. You get four scores, the working behind each, and a fix list in the order worth doing it.
Or email hello@oynk.co.uk