We tested 34 of Romania's best-known websites. Not one reached Google's "good" score

We wrote recently that a website's speed is measured rather than declared — and that the test is public, free and available to anyone. So we took the next step: we took 34 of Romania's best-known websites — banks, major retailers, telecoms, pharmacies, property portals, job boards — and ran exactly the same test on all of them: Lighthouse, Google's measurement tool, on the mobile profile, on the same day, under identical conditions.

The result, in short: not one of the 34 reached the 90-point threshold in the Lighthouse test — the one Google marks in green and calls "good". Not one.

One note before the numbers: this is not an indictment, and it is not a verdict on the companies behind these sites. It is a lab snapshot, taken on a single day, under identical conditions for all of them — and its limits are set out openly in the methodology.

The full ranking

The same numbers as a table — the score runs on a 0–100 scale, LCP is the time until the largest visible element in the first screen is rendered, which approximates the moment the main content becomes visible (Google's "good" threshold: 2.5 seconds), and CLS measures how much the page shifts while loading (threshold: 0.1). The first two rows are our own projects — reference points, not part of the 34-site sample:

A note on reading the table: a few points between two sites are not necessarily significant — lab measurements vary from one run to the next. The overall picture matters more than the exact order of any single position.

What the numbers show, beyond the ranking

• The median score is 44.5 — half of the 34 sites tested sit below 45 points, below the threshold at which Google colours a score orange at all (50).
• All 34 exceed the 2.5-second LCP threshold in the lab test. The best time measured was 4.3 seconds (BCR). Twenty-three of them go past 10 seconds.
• Banks lead the ranking — ING (65) and Banca Transilvania (64) hold the highest scores among the 34. A bank also sits near the bottom, though: under the conditions of the test, the BRD page had the longest display time in the study — 72.8 seconds on the first run, with a second run measuring 39.3 seconds. Even the better figure is more than 15 times the threshold.
• The two specialist electronics retailers in the test are among the five lowest results: Altex 30, Flanco 32 — a category in which customers often compare prices on a phone, standing in the shop.
• Visual stability is the hardest problem to spot: 7 sites exceed the 0.1 CLS threshold in the test — and an eighth, Vodafone, sits right on the line (0.0996, shown as 0.10 in the table). The highest figure is 1.09, more than ten times the threshold. At a CLS that high, elements carry on moving after the page appears — the kind of shift that can make you tap the wrong thing.

Why this happens

To be clear: these sites are built by capable people, inside large organisations. That is precisely the story. A large retail or banking site accumulates, year after year, layers that nobody can easily strip out again: legacy platforms, dozens of marketing and analytics scripts, cookie banners, A/B tests, chat widgets, personalisation. Every layer was added for a reason. Taken together, they are exactly the causes we wrote about: a site is not slow by accident, it is slow by construction — and at that scale, rebuilding is a matter of years rather than months.

The useful conclusion is not that somebody else got it wrong. Quite the opposite: this is where smaller businesses have the advantage. A business building its site today, from scratch, treating speed as an architecture decision rather than a finishing touch, can avoid the layers and compromises that large platforms accumulate over years — and then spend years repairing. It is one of the few places in the market where a small company can measurably score better on this test than a brand many times its size.

And us? The same measurements

The full methodology — so you can repeat the test

• The tool: Lighthouse 13.4, Google's performance test — the same one anyone can run at pagespeed.web.dev.
• The profile: simulated mobile (mid-range handset, slow 4G connection) — the standard conditions of Google's test, the same for every site.
• What we measured: we requested each site's main address (https://domain.ro/) and measured the page the site delivered, redirects included — on 28 July 2026, from the same location, on the same machine, our own site included. Seven sites delivered a page other than the bare domain: Banca Transilvania redirected to a campaign page (campanie.bancatransilvania.ro), Farmacia Tei to its shop at comenzi.farmaciatei.ro, Digi24 and OLX to their mobile versions (m.), BCR and ING to their personal-banking sections, and BestJobs to its job-listings page.
• Selection: we started with 36 sites; two (leroymerlin.ro and christiantour.ro) did not return a valid measurement in the test and were excluded.
• Our own sites: tangled-design.ro appears with the measurement taken after the optimisation carried out the same day (87 in the morning, 96 after the render-blocking CSS was removed — the final figure confirmed in two consecutive, identical runs). Every other site carries its morning-round figures.
• Fairness: for the 7 sites with extreme display times on the first run (LCP above 25 seconds) we ran the test twice and kept the result with the better score. At extreme timings, the variation between runs can be large — BRD, for example: 72.8 seconds, then 39.3.
• Limits, stated openly: this is a lab test — a controlled simulation, not the average of real visitors (Google measures that separately, from field data, and it can look different — sometimes better, sometimes worse). Scores change from one day to the next; yours today may differ from ours on 28 July. The test covers only the page delivered at the main address, not the whole site.
• Future edition: we will run each address several times and publish the median of those runs — for a ranking of precision, not only of overall shape.
• Right of reply: if you represent one of the sites tested and have better measurements or relevant context, write to us — we will gladly update the article, with the source included.

What to do with this

If you run a business, the question is not "why is eMAG slow", but "where do I sit in this ranking". You can find out in two minutes, free of charge: pagespeed.web.dev, your address, the mobile result. How to read the numbers — and above all why speed is settled at a site's foundations rather than at the end — is explained here. And if the result gives you pause and you would like an honest view of what could be done — repair or rebuild, with the costs in the open — tell us about your project. We measure before and after. Publicly, if you would like us to.