The weekly report across 66 sites shows 18.9k visits, up 2.72%, and 19.5k page views, up 0.52%. Median page load time came in at 5,680ms, up 133.06%. Two of those three numbers are the work of a single property, and the headline growth figure hides a decline almost everywhere else.
All of the growth came from one site
technologies.org went from roughly 1.6k visits to 4.6k, a gain of 187.5%. That is about 3,000 additional visits against total network growth of roughly 500. Take that one site out and the remaining 65 went from about 16.8k visits to about 14.3k, a drop near 15%. Page views excluding technologies.org fell around 17%.
So the network did not grow by 2.72% in any meaningful sense. One property absorbed a burst of traffic large enough to mask a broad softening across everything else.
Only three sites report performance data at all
Just three properties returned a page load time or Core Web Vitals this week: technologies.org, pho.tography.org, and analysis.org. Between them they account for about 7k of the 18.9k visits. The other 63 sites returned no beacon data.
That matters for reading the aggregate. Technologies.org at 6,291ms represents roughly two thirds of all measured sessions, so the network figure is essentially that one site’s number. Part of the 133% rise is a genuine regression and part of it is mix shift, since the slowest measured property tripled its share of the measured pool in the same week.
The genuine regression is clear enough on its own. Largest Contentful Paint at the 75th percentile moved from roughly 1,650ms to 6,056ms, up 267%. That is a move from good to poor in seven days. Cumulative Layout Shift went from about 0.15 to 0.22. Interaction to Next Paint held at 112ms, comfortably inside threshold, which points at delivery rather than script execution. On a WordPress property whose traffic tripled in the same window, the first thing to check is time to first byte and whether full page caching is doing anything.
Analysis.org is in similar territory at 5,167ms, with LCP at 3,200ms and CLS at 0.21. Pho.tography.org is the clean one: load time 1,228ms, down 28.84%, LCP 1,921ms, down 40.25%, and CLS at zero. Static output on Cloudflare Pages behaving exactly as it should.
The browser mix does not look human
Technologies.org reports Chrome at 3.02k, Unknown at 1.5k, and ChromeMobile at 40. Mobile is therefore under 1% of sessions on a site that tripled its traffic. Real editorial audiences run somewhere between 40% and 60% mobile. A third of the traffic carries no identifiable browser at all.
Analysis.org shows the same shape in miniature, with Unknown at 400 ahead of Chrome at 300.
Cloudflare’s analytics beacon fires on JavaScript execution, so this is not a plain crawler reading raw HTML. It suggests headless or scripted browsing, and it fits the performance story neatly. A script walking an entire archive misses cache on nearly every request, hits PHP and the database each time, and the origin slows down for everyone. The traffic gain and the load time collapse would then be the same event rather than two separate ones. Confirming that means looking at bot scores and user agent breakdowns in the main traffic dashboard, not in the analytics summary.
Everything below the top three
The clearest declines were marketanalysis.com at 700 visits, down 50%, 3v.org at 600, also down 50%, and technologyconference.com at 800, down 27.27%.
Marketanalysis.com is the only site in the top ten with more than one page view per visit, at roughly 1.4. Every other property sits at a flat 1.00. That is the signature of search landings that end where they start, and it is the largest structural opportunity in the table. Internal linking is the fix, and it costs nothing but editing time.
yellowfiction.com appears at 600 visits with no comparison figure, its first week in the report. defensemarket.com shows 600 visits and a 500% gain, which means it moved from roughly 100. At that volume the percentage is theatre.
One caveat on the smaller entries. Figures below 1k are rounded to the nearest hundred, which is why brandstoshop.com reads as exactly flat at 900 visits and 900 page views. Nothing holds that still. publishinghouse.org at 900 visits and up 350% is the one small-site gain large enough to survive the rounding.
Top ten by visits
| Site | Visits | Change | Page views | Load time |
|---|---|---|---|---|
| technologies.org | 4.6k | +187.5% | 4.7k | 6,291ms |
| pho.tography.org | 1.3k | +44.44% | 1.3k | 1,228ms |
| analysis.org | 1.1k | +120% | 1.1k | 5,167ms |
| brandstoshop.com | 900 | flat | 900 | no data |
| publishinghouse.org | 900 | +350% | 900 | no data |
| technologyconference.com | 800 | -27.27% | 800 | no data |
| marketanalysis.com | 700 | -50% | 1k | no data |
| 3v.org | 600 | -50% | 600 | no data |
| yellowfiction.com | 600 | first week | 600 | no data |
| defensemarket.com | 600 | +500% | 600 | no data |
What to fix before next week
Three things, in order. Establish whether the technologies.org traffic is an audience or a scraper, because the answer determines whether the LCP number is a caching problem or a content problem. Set explicit dimensions on images across technologies.org and analysis.org, since both sit above 0.2 on CLS and that is the cheapest of the three Core Web Vitals to repair. Then get the analytics beacon into the base templates of the other 63 properties, because a performance report covering 5% of the network is not a performance report.
Leave a Reply