React
React is a JavaScript library for building user interfaces. A page built with it usually arrives as a small HTML shell plus a bundle of code that constructs the interface in the browser after it loads, which is a different shape of page from one the server renders complete. You would meet the name in a job listing, a stack report or a devtools panel. This index identified it on 54 of the 5,294 sites measured, 1% of them, and those 54 sites behave differently from every other technology group here in one specific way: they start talking to third parties later.
The latest median first contact of the thirteen technologies here
The median site in this group made its first third-party request 989 ms after the page began loading. That is the latest median of the thirteen technologies in this index — no other group waits as long before its first outside contact.
A plausible reading is mechanical rather than moral. If the interface is assembled in the browser, the code that eventually inserts an analytics tag or an embed cannot run until the bundle has been fetched and evaluated, so the third-party requests queue behind the application itself. The index cannot confirm that chain of cause, but the pattern is consistent across the group: the earliest first contact recorded anywhere in it was 175 ms, and the named examples run from 503 ms on amazon.ca to 1,733 ms on globo.com.
Busy pages, ordinary company counts
The median site here issued 70 requests, second-highest of the thirteen technology groups, while contacting 7 companies — fifth by that measure, which is unremarkable. The two numbers are measuring different things. A client-rendered application makes many requests to itself: code chunks, data calls, images fetched after the first paint. Request volume is a property of the architecture; company count is a property of what the site chose to embed.
The median tracker count is 2, second-highest of the thirteen, so the group does skew towards commercially instrumented sites. Look at who is in it and that is not surprising.
Fifty-four sites is what could be identified, not what exists
React is on far more of the web than 54 sites. What this figure counts is the sites where the crawl could identify the library from the page as it was served, and a bundled, minified application does not always announce itself. Read the 54 as a sample of identifiable React deployments, not as market share; by site count the group ranks 13th of the 13 technologies in this index.
Who is in it: cloud and retail infrastructure (amazonaws.com at rank 9, amazonvideo.com at 119, primevideo.com at 765, amazon.ca at 647), large portals (qq.com at 87, nic.ru at 94), news and sport (globo.com at 416, espn.com at 435, dailymail.co.uk at 599 and dailymail.com at 698), and consumer platforms (ubi.com at 631, fiverr.com at 780).
Early contact on 52 of the 54
On 52 of the 54 sites, 96.3%, at least one company was contacted before the consent banner could have been answered — third-highest share of the thirteen groups. The median site had 5 companies inside that window.
The median site here also picked up a company after the window closed: the median count of companies appearing afterwards is 1, not 0. Combined with the late median start, this is a group whose third-party contact is spread across the page's life rather than concentrated in its first half-second.
globo.com is the extreme case in both directions: 36 companies contacted, the highest in these examples by a wide margin, but only 3 of them inside the consent window and a first contact at 1,733 ms. espn.com is the inverse — 8 companies, all 8 inside the window, first contact at 749 ms.
Consent tooling was the exception, not the rule
A consent tool was detected on 19 of the 54 sites. Only two products were seen alongside React at all: OneTrust on 12 sites and an IAB TCF consent tool on 7. Both dailymail domains were recorded with the IAB TCF string, espn.com and ubi.com with OneTrust.
The remaining 35 sites carried no detected banner, and the two highest-ranked examples are among them — amazonaws.com at rank 9 and qq.com at 87 — alongside primevideo.com and fiverr.com. A site can also run more than one banner product at once, so tool counts of this kind are never a clean division of the sites they came from.
| Site | Companies | Before consent | First contact |
|---|---|---|---|
| amazonaws.com | 6 | 4 | 1021 ms |
| qq.com | 4 | 1 | 1617 ms |
| nic.ru | 6 | 6 | 669 ms |
| amazonvideo.com | 5 | 3 | 1353 ms |
| globo.com | 36 | 3 | 1733 ms |
| espn.com | 8 | 8 | 749 ms |
| dailymail.co.uk | 12 | 4 | 1661 ms |
| ubi.com | 6 | 4 | 1162 ms |
| amazon.ca | 5 | 5 | 503 ms |
| dailymail.com | 12 | 4 | 1494 ms |
| primevideo.com | 5 | 5 | 790 ms |
| fiverr.com | 7 | 7 | 861 ms |
| Tool | Sites |
|---|---|
| OneTrust | 12 |
| IAB TCF consent tool | 7 |
Questions
Why is first contact so late on React sites?
The likely mechanism is ordering: the application bundle has to load and run before the code that inserts third-party tags exists. The measurement here is only the outcome — a median of 989 ms, the latest of the thirteen technology groups indexed.
Why were only 54 sites identified?
Because identification depends on the page leaving a recognisable trace, and a bundled application often does not. The 54 are the sites where the crawl could tell; they are a sample of React on the web, not a census of it.
Do React sites contact more companies than other sites?
Not particularly. The median is 7 companies, fifth among the thirteen technology groups here. What stands out is the request count — 70 at the median, second-highest — most of which is the application talking to itself.
Which example was the busiest?
globo.com, with 36 companies contacted. Three of them fell inside the consent window and first contact came at 1,733 ms, so the volume arrived late and spread out rather than all at once.