A year of capturing award-level sites, turned into numbers — and the numbers disagreed with almost everything I assumed about "modern" design.
I did not set out to build a dataset. I set out to stop guessing. Every time I admired a website I found myself saying vague things — "it feels premium," "it feels considered" — and I could never turn that feeling into anything I could reuse. So I started capturing the sites I admired and measuring them: the actual background colors, the real headline sizes, the letter-spacing, the corner radii, the number of sections. After 358 sites, the vague feelings had turned into numbers, and the numbers surprised me.
If you spend your days on Twitter/X, Dribbble and developer tools, you would assume the web went dark years ago. It did not. Of the sites I measured, 82% ship a light theme by default and only 18% are dark. Dark design is real and often gorgeous, but it is a genre — dev tools, crypto, studios — not the default of great design. I had confused my feed for the web.
The headline of a confident brand is loud, so I assumed it would be heavy. Wrong. Among the sites where I could read the H1 weight, the largest group uses a light or regular weight — more than the bold headlines. The confidence comes from scale, not thickness: a 60px median headline, sometimes far larger, set light and tracked tight. That single observation changed how I design.
Tech taught us the accent color is blue. But when I extracted the most saturated usable color from each palette, blue barely led — and the warm accents (orange and red) together clearly outnumbered it. The safest, most forgettable choice is exactly the one everyone defaults to. I wrote more about that in the color nobody expects.
Measuring did not make design mechanical. It did the opposite: it freed me from the adjectives that were quietly steering me toward the average. Once you know that great pages are usually light, that their headlines are usually light-weight, that their corners cluster around a 10px radius and that they rarely use pill buttons, you can break any of those rules on purpose instead of by accident.
That is the whole idea behind Norrly. Not to tell you what to make, but to hand you the measured decisions behind the sites you already admire, so your next page starts from a designed position instead of a default one. The full breakdown lives in the State of Web Design report, and every individual site is measured in the library.
I will keep writing these notes as the dataset grows. Each one takes a single measured fact and follows it somewhere useful. This first one is the summary; the rest go deeper.