Speedtests don’t work.

OK, this may sound a bit crazy, but there is some truth to this.

You see, there are no systems which can objectively test the raw speed of an internet (or network) connection. Every single test which can be used to measure the speed of a connection works in exactly the same way: throw increasing amounts of data at the connection until a certain percentage of packets don’t make it through. That data limit is returned as “the speed of the connection”. Of course, the speed of the connection isn’t what’s being measured by a test like this at all – it simply measures the available bandwidth at the moment the test is conducted. Under ideal circumstances, when the bandwidth use on the connection is negligible, this can provide a very good real-world result, but…

When the subscriber to a connection is utilizing more bandwidth than they realize (see my post “… Bandwidth Vampires“), there may not be much bandwidth left available, so any attempts to use the Internet at those times might be unreliable. If the download bandwidth is saturated, pages and images may load slowly, get interrupted, or return errors. If the upload bandwidth is saturated, the outgoing requests from the computer to the server may not even get sent out, resulting in all sorts of different errors. Unfortunately, when something appears wrong with the connection, many people think to run a SpeedTest to see how the connection is performing. If the connection is already under heavy use, the results will be much lower than expected, but that is not, in and of itself, evidence of a problem. The connection may well be working exactly as it’s supposed to, but the test results would be poor.

This does not mean that every instance of an internet connection performing below expectations is due to the connection being saturated; other factors, such as reduced signal level, Ethernet errors, or other network problems, could cause a connection to provide less throughput than it should, but actual incidences of these are relatively rare. If you’re in doubt, call us, and we can check all these things on your connection.

Now, if you are armed with the knowledge of how these tests work, you can make them work for you: if you find yourself under those circumstances, running and comparing test results while methodically shutting off devices one by one (or at least disabling their WiFi/Ethernet access) is a valid way of tracking down the device which is monopolizing the bandwidth. Sometimes it’s a smartphone, sometimes it’s a computer, sometimes it’s even an e-reader! The point is that no device can be ruled out by default – until you have ensured a particular device is not connected to the internet at all, you should be wary to assume it isn’t using up your bandwidth.