A speed test measures the path between one device and one server at one moment. Change any of those and the number changes. If you are about to complain to your provider or decide on an upgrade, it is worth spending five minutes getting a result that actually means something.
1. Use a cable if you possibly can
Wi-Fi is the largest source of misleading results. An Ethernet cable removes the wireless link entirely, so what you measure is the connection your provider delivers. If you have no Ethernet port, stand within a couple of metres of the router with a clear line of sight, and note that you are testing Wi-Fi rather than the line itself.
2. Pause everything else
Streaming on a TV, a console downloading an update, or a phone backing up photos will all take a share of the connection and depress your result. Pause them, and give the network a few seconds to settle before starting.
3. Close background applications on the test device
Cloud storage clients, system updates, and torrent software compete for bandwidth from the very machine you are testing on. Close them rather than trusting them to be idle.
4. Turn off VPNs and proxies
A VPN reroutes your traffic through another server, often in another country, and encrypts it along the way. Both effects reduce measured speed and increase latency. Test without it first to establish your baseline, then with it if you want to know the cost. Our guide on how VPNs affect speed covers what to expect.
5. Test more than once, at more than one time
Networks are busier in the evening, when whole neighbourhoods stream at once. A single slow result at 9pm proves very little. Run the test a few times across a day or two and look at the pattern rather than any individual figure. Our article on why results change during the day explains what those patterns mean.
6. Look past the download number
Download speed alone will not tell you why calls stutter or games lag. Read the ping and jitter figures, and the latency-under-load figures that MyRealSpeed reports during the transfer phases. A connection with plenty of speed but poor responsiveness feels slow in ways a headline number cannot capture. The Results Guide explains each one.
Recording results for your provider
If you are building a case that your service underperforms, keep a simple log: date, time, wired or wireless, and the four main figures. A dated series taken over a week is far more persuasive than one screenshot. Our guide to documenting poor performance sets out exactly what to collect.
Frequently asked questions
Does the browser matter?
Somewhat. Very old browsers and heavily extended ones can cap throughput. Test in an up-to-date browser with a clean window if results look implausibly low.
Can my device be the limit?
Yes. Older laptops and phones have slower network hardware and can plateau well below a fast plan, especially over Wi-Fi. If one device tests much lower than another on the same network, the device is the constraint.
Should I restart the router first?
It does no harm and occasionally clears a temporary fault, but do not do it mid-test. Restart, wait for everything to reconnect, then test.
