unctad.org speed test — 20 Oct 2026

Measured from the wfix.shop server on 20 October 2026 at 14:59 UTC. Grade D (69/100), with the screenshots, the timings and the fixes that matter.

https://unctad.org/

20 October 2026 at 14:59 UTC · Free test · report pt_d920443567d9

Re-test this page Test your own site

Grade

69/100

Performance58/100

Structure86/100

14 audit(s) passed, 2 need work

First byte (TTFB)1.86 smeasured from this server

Fully loaded3.26 sfirst byte + slowest asset

Requests26document + assets

Page weight23.9 KBcompressed transfer

HTML document23.9 KBbr

ProtocolHTTP/3cloudflare

Where the first byte went

DNS lookup 60 ms 3%
TCP connect 3 ms 0%
Server wait 1.79 s 97%

Measured on the connection that fetched the HTML. Every other request on the page waits for the end of this bar before it can start.

How this page compares

First byteslower than 97% of 4,999 pages tested here1.86 s against a median of 448 ms

Page weightlighter than 94% of 4,999 pages tested here23.9 KB against a median of 650.5 KB

Fully loadedslower than 84% of 4,999 pages tested here3.26 s against a median of 1.84 s

Requestssends more requests than 100% of 4,999 pages tested here26 against a median of 26

Every report published here is in this comparison, so it is a reference population of pages tested from one server - not a benchmark of the whole web.

What these numbers mean

The first byte took 1.86 s. Everything else waits for that number, so the server is where to start: caching, a slow query, or an overloaded host.
The page transfers 23.9 KB over 26 request(s).
The HTML itself arrives br-compressed.
It runs over HTTP/3, so the assets share a connection instead of opening one each.
The largest element took 2.80 s to paint, past the 2.5 s threshold: usually a hero image that is too big or a render-blocking script in front of it.
Layout stayed still (CLS 0.001), so nothing jumps under the reader's cursor.
The main thread was blocked for 174 ms in total, which is not enough to feel.

Fix first: Server response time (TTFB) - Cut server think time: page cache, query indexes, fewer synchronous calls to third parties inside the request. Connection setup is distance, so a nearer region or a CDN edge helps instead.

Read on one shared server, not a visitor's phone: use it to compare pages against each other, not as a promise about every connection. We fix what these tests find.

Screenshot

What the page looks like

Loaded in headless Chromium on this server. Click an image to open it full size.

Desktop screenshot of unctad.org — **Desktop** 1366 × 768 · 166.7 KB · page is 5529 px tall

Mobile screenshot of unctad.org — **Mobile** 390 × 844 · 87.2 KB · page is 6000 px tall

Browser render

The same URL loaded in a real browser, so these are the readings curl cannot produce. The grade above is still built from the server measurements.

Largest contentful paint2.80 sgood is under 2.5 s

Cumulative layout shift0.001under 0.1 is stable

Total blocking time174 mslong tasks during load

First contentful paint2.80 sfirst content on screen

Browser page weight6.91 MB123 request(s), scripts included

Failed requests5blocked or unreachable

Largest contentful paint2.80 s

Needs improvementgood under 2.5 s, poor over 4 s

Cumulative layout shift0.001

Goodgood under 0.1, poor over 0.25

Total blocking time174 ms

Goodgood under 200 ms, poor over 600 ms

Page weight by type

What the page is made of

Transfer size, compressed as sent. 26 request(s) in total, 23.9 KB across all of them.

Top issues

Server response time (TTFB)

high impact

The first byte arrived after 1857 ms. Of that, 1794 ms was the server thinking and 63 ms was connection setup. A first byte under 400 ms is what a visitor feels as instant.

Fix: Cut server think time: page cache, query indexes, fewer synchronous calls to third parties inside the request. Connection setup is distance, so a nearer region or a CDN edge helps instead.

Render-blocking resources

high impact

5 render-blocking resource(s) in the document head (0 script(s), 5 stylesheet(s)). The page cannot paint until each one arrives.

Fix: Add defer or async to scripts that do not change the first paint, inline the critical CSS, and load the rest with media="print" plus an onload swap.