Ask two people whether a website is good and you will usually get two opinions about colours. That conversation goes nowhere, because it treats a measurable thing as a matter of taste. Speed has thresholds. Accessibility has a standard. Touch targets have a minimum size in pixels. Line length has a researched optimum.
What follows is the standard we score against — eight dimensions, each with a published source. We have also flagged the places where the popular numbers are weaker than the people quoting them admit, because a standard you cannot defend is not a standard.
The bar is public, and most of the web misses it.
Two numbers frame everything else. On Core Web Vitals — Google's speed and stability measures — roughly 48% of mobile origins and 56% of desktop origins pass all three (Web Almanac 2025, from July 2025 CrUX data). And on accessibility, 94.8% of the top one million home pages have detectable WCAG failures, averaging 51 errors per page (WebAIM Million 2025).
Read those together and the conclusion is unusually cheerful: because most sites fail, simply meeting the published bar puts you ahead of the majority. "Good" is not table stakes. It is a position.
It also helps to know where sites fail. Speed is not one number — it is three, and they do not fail equally.
Share of mobile origins rated “good” on each Core Web Vital, and on all three at once (Web Almanac 2025, July 2025 CrUX). Load speed is the bottleneck that drags the combined figure down.
So if you only fix one thing on speed, fix how fast the main content appears. It is the measure most sites lose on.
The eight dimensions, and what "good" means on each.
Performance
How fast it feels on a phone, on real visitors' devices — not on your own laptop, which is always the fastest test you will ever run. Three things are measured: how soon the main content appears (aim for under 2.5 seconds), how quickly the page reacts to a tap, and whether the layout jumps about while loading. Load speed is where most sites lose. Developers call these LCP, INP and CLS.
Accessibility
Can everyone actually use it — including people with poor eyesight, or using a keyboard or screen reader. It is also the law in more places every year. The good news is that a handful of dull fixes cover most of it: text with enough contrast against its background, alt text on images, properly labelled form fields, no empty links or buttons. Low-contrast text alone affects around 79% of home pages — the single most common defect on the web.
Mobile & responsive
Most people will meet your business on a phone — roughly 62% of web traffic is mobile. That means buttons and links big enough to tap without missing (about the size of a fingertip, 44 pixels square is the working rule), text you do not have to pinch to read, and forms that work one-handed. A layout that technically loads on a phone is not the same as one someone has actually used on a phone.
Conversion & trust
Whether a visitor can tell what to do next, and believes you enough to do it. People form a first impression of a page in about 50 milliseconds (Lindgaard et al., 2006) — the judgement is made before a word has been read, and design measurably shapes it. In practice: an obvious next step on every page, real photographs of real work rather than stock, named proof near the point of decision, and a short form.
SEO & discoverability
Whether you can be found at all. The basics still matter — a clear page title, one main heading, a sitemap, and the structured data that earns richer search listings. For a local business the weighting is blunt: your Google Business Profile is the single biggest factor in local results and reviews are second (Whitespark's annual ranking-factors survey), and your primary business category is the most powerful setting you control.
Security & reliability
Whether the site is actually up, and safe to visit. HTTPS is the floor — browsers now warn on anything less. Uptime in plain terms: 99.9% still means about 8.8 hours offline a year, so it needs monitoring you actually see. Platform choice is a real risk decision — though the usual statistic is misread. WordPress makes up the overwhelming majority of the infected sites that cleanup firms handle, but it also powers a large share of the web and its owners are the ones who buy cleanups, so that figure is not a per-site risk rate. The narrower and more useful lesson: most compromises arrive through out-of-date plugins, so a platform with fewer moving parts simply has less to keep patched.
UX & design
Whether the page is comfortable to read. Body text at 16 pixels or larger, lines of about 50–75 characters (much longer and the eye loses its place), generous line spacing, and the important information first because people scan rather than read. None of this is decoration — it is the difference between a page that gets skimmed and one that gets skipped.
Content quality
Whether it says something useful, in plain language, and looks current. Aim for the reading level of a good newspaper, keep it honest, and skip the padding written for search engines rather than people. The cheapest credibility signal on the web is a page that was obviously updated this year.
We ran this on our own site, and it found things.
A standard you only point at other people is marketing. So we scored amkatechnologies.com against it, and fixed what it turned up. On Google's public mobile test the site now scores a median of 99 for performance across four runs — the individual results were 87, 96, 99 and 100, which is a fair illustration of how noisy a single test is — plus 100 for accessibility and 100 for SEO. Anyone can check that at pagespeed.web.dev.
More useful than the score is what the measuring exposed — because none of it was visible by looking at the site.
Our hero text was invisible until JavaScript loaded
The largest element on the homepage took 3.2 seconds to appear — but the download took no time at all. The cause was our own scroll animation: it hid content until a script ran. The text was there the whole time; the browser was not allowed to draw it. Desktop scored 99 throughout, so this was invisible unless you tested a phone.
Our own brand orange failed the contrast rule
Five places where saffron sat on light or navy backgrounds were below the legal minimum — including button arrows and a footer label. We now use slightly adjusted shades for small text and keep the brand colour for everything else.
Ten tap targets were too small
Measured on a phone-sized screen, ten links and buttons were under the size guideline and four were under the legal minimum — the smallest being the language switcher at 15×22 pixels.
Two honest caveats. Our traffic is not yet high enough for Google to hold real-visitor speed data on us, so those figures come from simulated tests rather than from what actual visitors experience — and the spread above is why you should run any speed test several times before believing it, ours included. And our own scorecard still lists things we have not finished.
Where the popular numbers do not hold up.
Anyone can quote a statistic at you. Fewer will tell you which ones are shaky — so here are the caveats on our own sources.
The speed-to-money figure is correlational. The most-cited study (Deloitte Digital and Google, "Milliseconds Make Millions", 2020) found a 0.1s mobile speed improvement associated with an 8.4% lift in retail conversions across roughly 30 million sessions. That is a large dataset and the direction has held up — but it is a correlation from 2020, not a controlled experiment, and it should not be quoted as a guarantee that shaving 100ms will pay for a rebuild.
"53% of users abandon sites slower than 3 seconds" is dated. It traces to a 2016/2017 DoubleClick study and is still repeated as though it were current. We do not use it.
Schema is not a ranking factor. Google is explicit about this. It earns richer search listings, which lifts click-through — a real benefit, just not the one it is usually sold as. Its effect on AI-generated answers is genuinely contested.
Shorter forms are not automatically better. Aggregate data shows a sharp drop between short and long forms, but controlled tests have found that removing fields can reduce conversions when those fields were building motivation or clarity. Test it; do not strip fields on faith.
"Mobile-first" needs a local check. Globally, mobile is about 62% of traffic. But on StatCounter's page-view metric, Malaysia, Singapore, and Indonesia are actually desktop-majority — an artefact of app-heavy usage. Design mobile-first anyway, but do not let anyone tell you your Malaysian B2B visitors are all on phones.
What this means if you own a website.
You do not need to score full marks on all eight. Most sites do not fail because of something exotic; they fail on the boring end — the contrast is too low, the page weighs four megabytes, nobody can tell what to do next, or the SSL certificate quietly expired and every visitor now gets a browser warning.
The useful exercise is not "is my site good?" but "which of these eight is actually costing me something?" For most firms the honest answer is one or two of them, and the fix is a morning's work rather than a rebuild. Anyone who responds to that question with a quote for a new site has not measured anything.
We score sites against this standard as a written review — including our own, which does not come out perfect. If you want to know where yours actually stands, ask.
Retrieved from https://amkatechnologies.com/insights/what-separates-a-good-website-from-an-average-one
Published 27 July 2026 · last updated 28 July 2026. AMKA Technologies Sdn Bhd, SSM 202301041763 (1535682-T).