What Is a Good Website Score? (And Which Numbers Actually Matter)

You ran your site through a checker, got a number, and now you’re trying to work out whether it’s bad.

Short answer: 90+ is genuinely good, 50–89 means there’s real work available, and under 50 means visitors are feeling it. But the number alone is close to useless without knowing which number it is — and most tools bury that distinction.

Here’s how to read your result properly.

First: which score are you even looking at?

“Website score” means at least four different things depending on what you ran:

Score	What it measures	Who sets the bar
Performance	How fast the page becomes usable	Google (Lighthouse)
SEO	Whether Google can crawl and understand it	Google (Lighthouse)
Accessibility	Whether people with disabilities can use it	Google (Lighthouse)
Overall / composite	A weighted blend of the above	Whoever built the tool

The first three are standardised — a 72 from one tool means roughly what a 72 means from another, because they’re reading the same Lighthouse audits.

The composite is not standardised. Every tool picks its own weights. That’s the score to trust least, and unfortunately it’s the one displayed largest.

What counts as “good”, by score

Performance

• 90–100 — genuinely fast. Rare. When I audited 24 real small business websites, exactly one hit this.
• 50–89 — the normal band. Usable, but visitors on phones are waiting longer than they should.
• 0–49 — people are leaving before your content appears. In that same audit, 9 of 24 sites were here.

Don’t chase 100. The gap between 90 and 100 is engineering effort with almost no commercial return. The gap between 40 and 80 is money.

SEO

• 90–100 — expected, not impressive. This is table stakes.
• Below 85 — something concrete is broken: a missing title, a blocked page, unreadable links.

Here’s the thing most people get backwards: a high SEO score does not mean you’ll rank. It means nothing technical is stopping you from ranking. Content, relevance and links decide the rest. Lighthouse can’t measure whether your page deserves to win.

Accessibility

• 90+ is a reasonable target and most of the gap is genuinely easy — contrast, labels, alt text.
• This one has legal weight in several countries, and it overlaps heavily with usability. Worth more attention than it usually gets.

The composite score problem

In my audit of 24 small business sites, the median overall score was 74 — a comfortable-sounding “room to grow”.

The median performance score was 62. One site took 35 seconds to show its main content on a mid-range phone, and still scored 52 overall, because near-perfect SEO and security pulled the average up.

That site was not a 52. It was a catastrophe with good metadata.

This applies to my own tool as much as anyone’s. So: read the categories, not the composite. If one category is on fire, that’s your answer regardless of what the big number says.

Where scores come from (and why yours keeps changing)

Most tools, mine included, use Google’s PageSpeed Insights. Two things surprise people:

1. It’s simulated, not your actual experience. Lighthouse loads your site on a throttled mid-range phone on a slow connection. That’s deliberate — it approximates a real customer, not you on office wifi with everything cached. Your site feeling fast to you means very little.

2. Scores fluctuate between runs. Running the same page twice can differ by 10+ points, because network conditions and server response vary. Testing this article’s own site three times in a row, I got 98, 83, 98 — the 83 was a single slow resource fetch, not a real regression.

So: run it 2–3 times and take the median. A single run is a data point, not a verdict. Any tool presenting one run as definitive truth is overselling.

What actually moves the number

From the 24 sites I audited, ranked by how often each was the problem:

1. Images. The single most common culprit in sites I’ve actually opened up. Oversized, uncompressed, wrong dimensions, still in JPEG/PNG. Converting to WebP or AVIF at correct dimensions typically cuts load time substantially with zero visible design change.

2. Too much JavaScript. Every script runs before your customer sees anything. Plugins, chat widgets, tracking pixels, A/B tools — each one is a tax. If you can’t name what a script does, it’s a candidate for deletion.

3. Third-party scripts specifically. A recent, real example from this site: adding Google Analytics cost my own homepage ~17 performance points, because the analytics script was competing for bandwidth with the hero image. Fixed by loading analytics after the page finishes. Same tools, same content, 17 points — purely from when something loaded.

4. Render-blocking CSS and fonts. Anything the browser must fetch before it can draw. Custom fonts are the usual offender.

5. Hosting and server response. Real, but usually smaller than people hope. Switching hosts rarely fixes a site that’s slow for reasons 1–4.

Notice that none of the top four require a rebuild. Most slow sites are fixable without being replaced — and you should be suspicious of anyone who quotes you a rebuild before showing what a cleanup would achieve.

So what should you actually do?

1. Run your site 2–3 times, take the median.
2. Ignore the composite. Look at Performance and SEO separately.
3. If Performance is under 50, start with images. It’s the cheapest fix with the biggest return.
4. If SEO is under 85, find the specific failing check — it’s usually one concrete, fixable thing.
5. Re-test after each change so you know what actually worked.

You can get all four category scores, and the specific list of what passed and failed, from my free Website Report Card — about a minute, no email, no signup.

And if the list is longer than you want to deal with, message me on Fiverr. I’ll tell you honestly which items are worth your money and which ones you can safely ignore — that second list is usually longer than people expect.