Measuring basic website health across a region: the method and its limits
How to test thousands of business websites for three basic defects, what our scan of one Dutch region found, and what a scan like this cannot tell you.
The short answer
There is a small set of website defects that can be measured from the outside, cheaply, across thousands of sites at once, because each one is a single question about the HTML that comes back. We ran that measurement across one region of the Netherlands in August 2026 and published the method here alongside the result, because the method is the part you can reuse and check.
The result, in one line: of 3,059 business websites that responded, 475 carried at least one of three basic defects. That is 15.5 percent.
Where the addresses came from
The list was not bought and it is not a sector list. It was taken from OpenStreetMap, the open map database, by selecting mapped businesses that record a website address, in three areas:
| Area | Addresses |
|---|---|
| Limburg-Zuid | 1,630 |
| Limburg-Noord | 1,214 |
| Zuidoost-Brabant | 830 |
| Total scanned | 3,674 |
Of the 3,674 addresses, 615 did not respond to our request: a connection error, a refusal, or a status other than OK. That is 16.7 percent, and those sites are excluded from every figure below rather than counted as defective.
What was measured, and how
Each reachable page was fetched once and the returned HTML was tested three times. The tests are deliberately crude, because a crude test that is the same for every site is more honest than a subtle one that drifts.
| Signal | The test | Found | Share of 3,059 |
|---|---|---|---|
| No HTTPS | The site was reachable only over http://, so a browser shows it as not secure | 285 | 9.3% |
| Not mobile-ready | No <meta name="viewport"> in the HTML, so the page renders at desktop width on a phone | 171 | 5.6% |
| Markup from the 2000s | A <font>, <marquee>, <blink> or <center> element, or a reference to Flash | 121 | 4.0% |
| At least one of the three | 475 | 15.5% | |
| Two or more | 92 | 3.0% |
The three shares do not add up to 15.5 percent, and that is the point of the last two rows: 92 sites carry more than one defect, so the union is smaller than the sum.
What this method cannot tell you
This is the section that decides whether a number like 15.5 percent is worth anything.
- It does not measure quality. A site can pass all three tests and still be slow, confusing, inaccessible or wrong about what the business does.
- It does not measure the business. A shop with an old website can be thriving; a modern site can belong to a company with no customers.
- It is one region and one moment. August 2026, three areas in the south-east of the Netherlands. It is not a statement about the country, and it is not a trend, because it was measured once.
- It sees only mapped businesses with a listed website. Anything outside OpenStreetMap, or inside it without a website tag, is not in the sample.
- It did not open a browser. Every judgement comes from one plain request. A site that renders its content with scripts could be tested wrongly by the viewport check, and we did not check how often that happened.
- The 615 unreachable sites are unexamined. They may include the worst sites in the sample, or none of them. Their absence is the largest single limitation of the figure.
Why we published the method rather than the leads
The same scan can be used to build a call list. That is what this kind of tooling is usually for, and it is why a percentage from a supplier deserves the questions above: the measurement and the sales motive often share a spreadsheet.
What is useful to a reader is not our list but the three checks, which cost nothing to repeat on one site. The defects are also cheap to prevent: a certificate is issued automatically by any modern platform, a viewport tag is one line, and the 2000s markup is a sign that nothing has been touched in a long time.
What this means for you
Run the three checks on your own site in a minute: open it over http:// and watch whether it switches to https://, view the source and search for viewport, then search the same source for font, marquee and swf. Three answers. If all three are clean, this scan would not have found your site, and that is the entire claim the scan makes.
Frequently asked questions
Why exclude the sites that did not respond?
Because a failed request from a script is not evidence that a site is down. It can be a firewall, a bot filter, a temporary outage, a domain that has moved, or a rate limit hit by our own scan. Establishing which would mean opening each one in a browser, which we did not do. Counting all 615 as broken would have produced a much worse and much less honest number, so they are outside the denominator and named separately.
How were the addresses chosen?
From OpenStreetMap, by area rather than by sector: businesses with a website tag in Limburg-Zuid, Limburg-Noord and Zuidoost-Brabant. That gives a sample of businesses that are mapped and that list a website, which is not the same as all businesses in the region. Any business that is not in OpenStreetMap, or that is there without a website, is invisible to this method.
What does no viewport meta tag actually mean for a visitor?
The page renders at desktop width on a phone, so the visitor arrives at a zoomed-out page and has to pinch and scroll to read anything. It is the single clearest sign that a site was built before mobile mattered, or built without testing on a phone. The tag itself is one line of HTML; its absence is not the problem but the symptom.
Can we run the same check on our own site?
Yes, and it takes a minute without any tooling. Open your site over http:// and see whether the address bar switches to https:// on its own. View the page source and search for the word viewport. Search the same source for the words font, marquee and swf. Three answers, and they are the same three this scan counted.
Sources
- OpenStreetMap (accessed 2026-09-14)
- MDN Web Docs: <meta name="viewport"> HTML attribute value (accessed 2026-09-14)