How we measure broadband — and why we built this

Who runs the site

BroadbandSpeedChecker.net is run by a small independent team. We're not an ISP, an affiliate network, or a hardware vendor — which means we have no incentive to flatter any particular connection type or brand. The site is built and maintained as a side project, with revenue coming from non-intrusive ads only.

You can reach us at [email protected] — we read every message.

How the speed test actually works

Under the hood we use Cloudflare's measurement endpoints. When you press Start, your browser opens connections to the Cloudflare edge data center closest to you (out of more than 300 worldwide) and runs these steps:

1. Latency & jitter: a series of small round-trip probes measures how long packets take to travel from your device to the edge and back, and how much that timing varies between probes.
2. Download: your browser fetches test files from the edge, stepping up from 100 kB and 1 MB to 10 MB and, on fast lines, 25 MB, 100 MB and 250 MB, taking several samples at each size. It stops stepping up once a single transfer takes longer than about a second, so slower connections never download the large files. Your download speed is the 90th percentile of the measured transfers.
3. Upload: the same process in reverse, pushing files from 100 kB up to 50 MB back to the edge. Download and upload rounds alternate, so a brief slowdown in one moment can't skew the whole result.
4. Ping under load: while the download and upload run, the test keeps pinging the edge. The difference from your idle ping is the lag your line adds when it is busy, often called bufferbloat. We grade it from A+ (under 5 ms added) to F (over 400 ms added).
5. Analysis: results are calculated in your browser. No result ever touches our servers. Tick “Run 3 tests” and we report the median of three runs, which smooths out one-off blips.

What we don't measure (and why)

A speed test is only ever a snapshot of one path through the internet at one moment. We're explicit about what we deliberately don't report:

• Packet loss. Measuring real packet loss requires UDP probes via TURN endpoints. Cloudflare blocks cross-origin access to those for any site outside their own domain, so we can't get an honest number from the browser. Rather than show a fake or unreliable figure, we leave it out.
• Every destination at once. Ping under load is measured to the same Cloudflare edge as everything else. It shows how much lag your own line adds when it is busy (bufferbloat), but not how a particular game or call server will behave.
• Routing quality to specific destinations. Your latency to a Cloudflare edge isn't the same as your latency to a specific gaming server or call peer. The number we show is a fair baseline for “is your last mile healthy?” but not a substitute for in-app latency.

How accurate is it?

We use the same measurement engine, the same sequence of test files and the same method (90th-percentile throughput, median latency) as Cloudflare's own speed.cloudflare.com, against the same edge servers. Results from the two should normally be close, though no two runs are ever identical.

The test adapts to your line. A typical test takes about 30 seconds. That's deliberate: shorter tests take too few samples and swing by 20% or more from run to run. On gigabit lines the test moves more data (a few hundred megabytes) because short transfers can't fill a fast line and would read low. Browser tests also have a ceiling: on multi-gigabit lines your device's CPU and browser often limit the result before your connection does.

What we don't do

• No account, no email. The site is entirely anonymous to us.
• No server-side history. Your “Recent tests” list lives in your browser's localStorage, on your device only.
• No selling of data. We have no data to sell.
• No fingerprinting or tracking scripts beyond what's needed for ads (where displayed) and basic server logs.

Full detail is in our Privacy Policy and Cookie Policy.

Editorial standards

Our guides are written and reviewed in-house. We aim for plain English, and we'd rather say “it depends” than invent a confident-sounding generic answer. When we cite figures — for example, Ofcom UK statistics or FCC benchmarks — we link to the source so you can check it.

We update each guide when the underlying facts change (new ISP technology, new recommendations from regulators, new household usage patterns). The “Updated” date at the top of every article is the date of the most recent meaningful revision, not just a cosmetic refresh.

How the site is built

The site is hand-written in plain HTML, CSS and JavaScript, bundled with Vite and served as a small static site. There is no tracking framework, no analytics SDK, and no third-party JavaScript beyond the measurement engine itself. Page weight on the homepage is well under 300 KB.

Get in touch

Spotted a bug, a misleading number, or a guide that's gone out of date? Tell us: [email protected]. Press and partnership enquiries are welcome at the same address.