How SpeedFibre measures your connection
What each number means, how it is measured, and what a browser test cannot tell you.
Last reviewed: 2026-11-03. Written by Zakaria Z., fibre optic engineer (FTTH).
What the test measures
| Result | How it is measured | Unit |
|---|---|---|
| Download | Files of growing size (100 kB to 250 MB) are downloaded until a single request lasts about 2 seconds. The speed shown is the rate sustained over the last seconds of transfer, so the ramp-up at the start does not count. | Mbps |
| Upload | The same method in the other direction, with files from 100 kB to 50 MB. | Mbps |
| Ping (idle latency) | 20 small requests on an idle line, before the transfers start. We show the median. | ms |
| Jitter | The average difference between consecutive pings of that idle series. | ms |
| Latency under load | Small requests sent every 0.4 seconds while the download, then the upload, fills the line. We show the higher of the two medians and how much it adds to the idle ping. | ms |
The engine is Cloudflare's open-source speed test library, the same one used by speed.cloudflare.com.
Which server you test against
Your browser connects to the nearest data centre of Cloudflare's global network. The city is shown under the test as "Server". Which data centre you reach depends on how your provider routes traffic, not on us: two people in the same city can reach different data centres.
If Cloudflare cannot be reached, the test falls back to our own server, which has a single location. Results from that fallback are usually lower when you are far from it.
What a browser test cannot measure
- Packet loss. Browsers only send web traffic over TCP, which retransmits lost packets silently. Measuring loss needs a special server and protocol that this site does not run, so we do not show a packet-loss figure.
- Traceroute. A browser cannot send the raw network packets a traceroute needs. Use `tracert` (Windows) or `traceroute` (macOS, Linux) on your computer.
- Your Wi-Fi link rate. A web page cannot read your Wi-Fi signal. Our Wi-Fi or line diagnosis compares a test near the router with one where you sit instead.
- What your device cannot handle. An old laptop, a 100 Mbps Ethernet port or a busy browser tab caps the result.
Why results differ between speed tests
Each tool uses different servers, a different number of parallel connections, a different duration and a different way to summarise the data (peak, average or median). Network headers also take about 5 % of a line, so a 1 Gbps plan typically shows around 940 Mbps. For a fair comparison, test over a cable, at the same time, with nothing else running, and run each test more than once.
How the public statistics are built
The Internet Speed Index only uses results of tests run on this site:
- a country or provider is shown only when it has at least 30 results in the period;
- one device counts once per day (its latest test), so a single visitor cannot flood an average;
- we publish medians, which resist extreme values better than averages;
- no IP address is stored. See the privacy policy.
These figures describe the people who chose to test on SpeedFibre. They are not a representative survey of a country.
Get the most accurate result
- Use an Ethernet cable, close to the router.
- Pause downloads, backups and streaming on every device.
- Run the test two or three times, at different hours.
- Note the latency under load, not only the download speed.
Frequently asked questions
Is a browser speed test accurate?
It measures real data transfers, so it is a valid measurement of what your browser gets from the test server. It cannot see beyond that path: Wi-Fi limits, a weak device or a distant server lower the result.
Why does my result change between two tests?
The line is shared with other traffic, Wi-Fi conditions change and the test may reach a different server. Small variations are normal; look at the trend over several tests.
Do you measure packet loss?
No. A browser cannot measure it reliably over normal web traffic, so we do not show an estimate.