How I score a website on design, content and speed

Every site I quote on gets a score out of 100 first: design 50, content 30, speed 20. Design is judged against the norms of the trade, content against whether the offer and the work are legible, speed against Google PageSpeed on mobile. Below is the whole rubric, published.

I publish it for two reasons. First, a score nobody can inspect is just an opinion with a number attached. Second, most of the studios I write to already suspect what is wrong with their site; the rubric tells them which part, and how much it costs them.

What exactly does the rubric measure?

Three axes, weighted 50 / 30 / 20, each broken into sub-criteria so the number is reproducible rather than a feeling. The weights come from what actually changes an enquiry rate for a visual business, in that order.

Design, 50 points

• Typography, 15. Typeface choice appropriate to the trade, a real type scale rather than four random sizes, line length in the 55 to 75 character range, line height that survives on mobile, and no system-font fallback flashing in on load.
• Hierarchy and layout craft, 15. Does one thing dominate each screen. Consistent grid and margins, deliberate white space, alignment that holds at 390 px and at 1920 px. This is where template sites lose the most, because a theme gives you spacing that was designed for somebody else's content.
• Imagery, 12. Quality and consistency of the photography, crop discipline, and whether the best work appears above the fold or three scrolls down. Mixed-quality galleries score worse than small ones.
• Detail and consistency, 8. Button and form states, hover behaviour, focus rings, icon set coherence, favicon, 404 page, the small surfaces that reveal whether anyone finished the job.

Content, 30 points

• Offer clarity, 10. Within five seconds: what this practice does, for whom, where. A studio name over a slideshow scores near zero here, no matter how good the slideshow is.
• Work shown, 10. Number of projects presented properly, with location, scope, year and a caption that says something. A site with six complete project pages outscores one with forty thumbnails and no text.
• Proof and specificity, 6. Named clients where permitted, real awards and press with links, team and credentials, service and process described in the trade's own vocabulary.
• Writing, 4. Sentences that a client would recognise as coming from this practice. Generic luxury copy is penalised here, and it is the easiest four points to recover.

Speed, 20 points

• Mobile performance score, 12. Straight from PageSpeed Insights, mobile tab, on the homepage and one project page.
• Largest Contentful Paint, 4. Google's threshold is 2.5 seconds or under; see the Core Web Vitals reference. Portfolio sites fail this more than any other metric, because the hero is usually a full-bleed photograph nobody compressed.
• Stability and interaction, 4. Cumulative Layout Shift at 0.1 or below, Interaction to Next Paint at 200 ms or below.

Why is design weighted at half?

Because in the built environment the website is a sample of the work. An architect, an interior designer or a boutique agency is selling judgment about how things look and how they are made, and the site is the first artefact of theirs a client ever handles. A slow site loses patience; an ugly site loses the argument.

Content takes a third because clarity converts, and speed takes the remaining fifth because it is the cheapest thing to fix. I can move a mobile performance score by forty points in a day of image work. I cannot move a design score by forty points without rebuilding the surface.

That ordering also stops the rubric from rewarding the wrong site. A stripped-back page with no photography will pass every technical test and still be worthless to a studio whose entire case rests on images. I wrote more about what separates the top of this field in what makes an architecture website win awards.

How do you judge design without judging taste?

By judging against the norms of the trade rather than a single aesthetic. A bold, service-led home builder site is scored as a builder site: does it make the offer, the service area and the phone number unmissable, does it show completed homes with specifications. Judged as a gallery it would fail, and the failure would be mine, not the builder's. That distinction is the whole reason the design axis has four named sub-criteria instead of one impression, and it is why the builder advice in home builder websites that book site visits looks nothing like the advice I give a design studio.

The same applies to property. A boutique agency's brand pages and its listing pages are doing two different jobs and score differently, which I set out in listing pages versus brand pages.

Within a trade, the calibration comes from volume. I have run this rubric across hundreds of studio, builder and agency sites in the tool I built for it, which means a 62 in one trade means roughly what a 62 means in another. I will publish the aggregate distributions once the sample per trade is large enough to be worth reading; until then, the rubric is the artefact and the numbers stay unpublished.

What do the score bands mean in practice?

Four bands, and they map onto what I recommend rather than onto praise.

• 80 and above, crafted. Somebody with an eye made this and it works. My honest answer is usually that a rebuild would change nothing. Occasionally there is a speed or structured-data job worth doing.
• 60 to 79, competent. The site is fine and the practice has outgrown it. This is where an overhaul earns its money: keep the URLs and the search equity, rebuild the surface.
• 40 to 59, template. A theme doing its best. Everything is present, nothing is specific, and the work is being flattened by somebody else's layout. A new site changes the enquiry rate here.
• Below 40, neglected. Broken layouts on mobile, missing project pages, PageSpeed in the twenties. Cheapest to improve and the most likely to be actively costing enquiries.

A score on its own is not a sales argument, so the proposal always names the three highest-value fixes and what each is worth in points. If those fixes total less than fifteen points, I say so and do not quote. Cost expectations for the work that follows are set out in what an architecture website costs in 2026.

What does the rubric deliberately ignore?

Novelty. Motion for its own sake, cursor tricks, scroll-jacking and WebGL do not earn points, and heavy versions of them lose points through the speed axis. Interaction only counts when it carries information.

It also ignores platform. A hand-written Next.js build and a well-handled Squarespace site can both score in the seventies. What it does not ignore is the consequence of the platform, and that shows up as page weight and layout rigidity soon enough.

I hold my own work to the same test. The opening film on this site is 288 hand-sequenced frames drawn on a canvas rather than a video element, and every journal header is compressed to WebP under 180 KB, because a hero that misses the 2.5 second threshold would cost me speed points I ask other people to earn.

Can I score my own site with this?

Yes, and it takes about twenty minutes. Run PageSpeed Insights on your homepage and one project page, mobile tab, and take the average for the twelve-point sub-criterion. Then open your site on a phone you have not cleared the cache on, and answer the content questions out loud: what do we do, for whom, where, and can a stranger see three complete projects in under a minute.

Design is the axis you cannot mark yourself, so ask someone whose eye you trust and give them the four sub-criteria rather than asking whether they like it. If your total lands in the sixties, the site is not the problem you think it is; you have most likely outgrown it. If it lands in the forties, the template is holding the work back.

The rubric is not a secret and it is not the product. It exists so that the recommendation which follows can be argued with. More on how I work and what I build is on the page about me and the way the work is done.