Short version
Among the 19 agencies in our benchmark, the #1 results in Belgium and the Netherlands have the lowest lab Performance scores in their respective countries. Lab performance does not consistently track better rankings in this sample; Page Authority has a stronger association. These observations do not establish cause.
Sort our web agency benchmark by Google position and something surprising appears. Belgium’s #1 agency, Inspira, has the country’s lowest lab Performance score: 27 out of 100. The Netherlands’ #1, Yourhosting, has the lowest score in its country: 28 out of 100. Their lab LCP values are 19.4 and 21.4 seconds. If speed matters for rankings, how can sites with scores like these come first? I investigated after publishing the benchmark.
What I researched
The September 10, 2026 benchmark covers 19 web agencies: businesses in the Belgian and Dutch top 10 for “website laten maken” (having a website built) that build websites themselves. The comparison platform at Dutch position 6 is excluded. I tested each homepage three times with the mobile settings in the PageSpeed Insights API and took the median for each metric.
One day later, I added field data from Google’s Chrome UX Report (CrUX): LCP distributions from an eligible subset of Chrome users, aggregated over 28 days. I also collected Moz link data (Domain Authority, Page Authority, and linking root domains), a fresh SERP snapshot with each ranking URL, and an on-page analysis of the ten highest-ranking URLs covering titles, h1 headings, word counts, prices, and structured data.
This is a small observational study, not an experiment: 19 agencies, one query, and measurements from September 10–11. The SERP was retrieved through a data-center connection. The correlations are indicative and cannot establish cause. There is also a scope mismatch to keep in mind: the lab tests measure homepages, while many ranking URLs are other pages. URL-level CrUX data and origin-level aggregates must not be treated as interchangeable measurements of those ranking pages.
The full table: speed next to authority
The table below sorts agencies by Google position within each country rather than by lab score. Lab values come from September 10; Moz and CrUX data were retrieved on September 11, 2026. Twelve agencies had no usable current CrUX observation, so they receive a dash. Missing data does not tell us why no result was available for a URL or origin, or whether it was fast or slow. CrUX LCP (p75) is the value at or below which 75% of measured page loads fall, not a measurement of every visitor’s experience.
| Agency | Lab Performance | Lab LCP | CrUX LCP (p75) | DA | PA |
|---|---|---|---|---|---|
| Inspira (BE #1) | 27 | 19.4 s | — | 22 | 29 |
| Conversal (BE #2) | 70 | 18.2 s | 1.09 s | 33 | 43 |
| Kreatix (BE #3) | 41 | 7.5 s | — | 25 | 32 |
| Green Bananas (BE #4) | 35 | 18.8 s | 1.63 s | 23 | 30 |
| KMO Shops (BE #5) | 47 | 4.8 s | 2.51 s | 32 | 24 |
| Adworkx (BE #6) | 43 | 21.2 s | — | 12 | 24 |
| Zenjoy (BE #7) | 82 | 3.0 s | — | 24 | 26 |
| Swift Marketing (BE #8) | 38 | 26.4 s | — | 13 | 24 |
| Riseweb (BE #9) | 59 | 11.7 s | — | 10 | 13 |
| WebsiteDeal (BE #10) | 59 | 9.4 s | — | 1 | 33 |
| Yourhosting (NL #1) | 28 | 21.4 s | 2.08 s | 70 | 49 |
| WJ Webdesign (NL #2) | 96 | 1.7 s | — | 15 | 39 |
| Webstijn (NL #3) | 51 | 15.4 s | — | 15 | 15 |
| VrijdagOnline (NL #4) | 43 | 12.2 s | — | 25 | 40 |
| Websteen (NL #5) | 52 | 11.0 s | — | 19 | 25 |
| 2manydots (NL #7) | 39 | 14.5 s | 2.19 s | 26 | 38 |
| Stuurlui (NL #8) | 92 | 3.2 s | 1.21 s | 35 | 37 |
| Scrolla (NL #9) | 46 | 10.6 s | 1.31 s | 24 | 25 |
| Creative Corner (NL #10) | 62 | 9.9 s | — | 57 | 25 |
Read DA and PA as Moz estimates based on link profiles, not as Google scores. They let us compare a different set of characteristics with lab performance. A correlation involving PA does not mean Google uses Moz’s score as a ranking factor.
Speed and position in this sample
The Spearman rank correlations show no consistent association between better loading performance and better rankings in this sample. The lab comparison covers 19 agencies; the current CrUX comparison covers only seven. A positive ρ means higher values tend to go with higher, and therefore worse, position numbers.
| Factor | ρ | How to read it |
|---|---|---|
| PA of the ranking URL | −0.52 | Higher PA, better position — strongest association |
| DA of the domain | −0.20 | Weak relationship |
| Linking root domains (log) | −0.14 | Almost no relationship |
| Lab score (Performance) | +0.32 | Higher score, slightly worse position |
| Lab LCP | −0.28 | Weak association, opposite to the expected direction |
| CrUX LCP (p75) | −0.07 | Near-zero association (n=7) |
Higher lab scores tended to go with slightly worse positions, while CrUX LCP (p75) had a near-zero association with position. Neither result proves that speed does not matter. Google uses Core Web Vitals, and LCP alone is not a complete assessment of them. The separate History API sweep returned records for 12 origins, ten with numeric LCP values; those records are not the seven current observations used in this correlation.
What does track: link equity on the page
The ranking URL’s Page Authority has the strongest association in the dataset, at ρ −0.52. The association with the domain’s DA is weaker, at −0.20. That points to a question worth investigating: what is different about the pages that rank, not just the domains behind them?
How the Dutch #1 built its authority
Yourhosting is a hosting platform with DA 70 and 46,561 linking root domains. Its anchor data includes 11,285 linking domains with empty anchor text, 10,556 with “contact,” and 9,780 with “algemene voorwaarden” (terms and conditions). The research identified footer links from hosted customer sites as part of that profile. This reflects the scale of a hosting business, which a small independent agency is unlikely to reproduce. It does not establish how much those links contributed to the ranking.
How smaller agencies get links
The smaller agencies have different link profiles. WJ Webdesign (DA 15, NL #2) has 70 linking domains using the anchor “website laten maken.” Conversal (DA 33, BE #2) has 144 using the same anchor. Those links coexist with high rankings, but the data does not isolate their effect. I would not copy an exact-match anchor pattern as a ranking tactic; links arranged to manipulate rankings carry a risk under Google’s spam policies.
Inspira, the Belgian #1 with DA 22, has another pattern: brand anchors from customer sites and topical links to the pricing blog that ranks first.
The right URL for the query
Only three of the nineteen agencies rank with their homepage. Eight rank with a landing page, four with a blog post, three with a service page, and one with an FAQ page. For this query, the ranking page is usually a dedicated answer rather than a general introduction to the agency.
A small agency’s high-ranking pricing guide
Inspira is a useful example. Its domain has DA 22 and 439 linking root domains, but its ranking URL directly addresses a likely question behind the search: what does a website cost? The blog post’s title closely matches that question, and the page discusses prices in detail. It illustrates a pattern across the top 10: prominent pages address the searcher’s needs directly.
What the winners do on the page itself
The on-page analysis of the ten highest-ranking URLs (top 5 per country) shows a strikingly uniform pattern:
- All ten include the exact search term in the title and h1.
- Nine of the ten show prices on the page, and nine of the twenty search results included pricing information in the snippet.
- Nine of the ten have FAQ structured data alongside visible questions.
- Four of the Dutch top 10 have review stars in the snippet (Stuurlui 4.8 from 124 reviews, Websteen 4.9 from 74), a feature almost absent from the Belgian results in this snapshot.
These observations concern content and presentation, not loading speed. They give us topics to investigate, but the snapshot does not show whether any individual feature increased rankings or clicks.
What this means for your website
Start by reviewing field data where available, then use lab tests to investigate performance problems. Good Core Web Vitals matter, but they do not guarantee rankings. Also ask which page best answers the query. In this sample, that is usually a landing page or blog post rather than a homepage. When you promote your work, link to the page that is most useful in context instead of sending every link to the homepage.
I apply the same scrutiny to our site. With DA 18 and 306 linking root domains, we are not on page 1 of this SERP yet. This research is a working hypothesis for our own decisions, not a success story we have already proved. It gives us reasons to improve the pages that answer the query rather than chase another Lighthouse point solely for SEO.
For a practical approach to your own site, read our guide to higher rankings. The difference between lab tests and real-user performance data is covered in lab versus field. We look more closely at how smaller agencies compete in small agency, big competitors.