2026-09-19 · IPV4 Scanner
You run a speed test and get 240 Mbps. You run it again and get 90. Nothing changed in between, and now you have no idea what your connection actually does.
This is normal. A speed test measures an entire path — your device, your Wi-Fi, your router, your provider's network, the route to the test server and the server itself — and reports the narrowest point at that moment. Change any link and the number changes.
Start here, because it explains most variation people blame on their provider.
Wi-Fi is shared radio. Your neighbours' networks, a microwave, a Bluetooth device and the wall between you and the router all reduce throughput, and none of them are constant. The 2.4 GHz band is particularly crowded, and while 5 GHz is faster and cleaner, it travels less well through walls.
The test is simple. Run a speed test over Wi-Fi, then plug into the router with an Ethernet cable and run it again. If the wired result is close to what you pay for and the wireless one is not, your connection is fine and your Wi-Fi is the constraint. That is a different problem with a different fix, and no package upgrade addresses it.
A speed test measures the path to one specific server. Pick a nearby, well-connected one and you measure close to your line's capability. Pick a distant or busy one and you measure the route rather than your connection.
This is why results differ between testing services, and why the same service can disagree with itself after choosing a different server.
An older laptop with an old Wi-Fi chipset cannot receive at modern speeds no matter what the line delivers. A 100 Mbps Ethernet port — still common on older equipment — caps everything at 100 regardless of what you pay for. One elderly switch or a damaged cable anywhere in the chain limits everything behind it.
Browsers add their own ceiling. A speed test running in a tab with forty other tabs open, on a machine that is thermally throttling, measures your computer as much as your connection.
Anything else using the connection during the test competes with it. A cloud backup, a game console updating in standby, a phone syncing photos, someone else streaming. Some of this is invisible — consoles and computers download updates while apparently asleep.
For a clean reading, test when nothing else is running. For a realistic reading, test during normal use. Both are legitimate; they answer different questions.
Evening congestion is real, particularly on shared-medium connections like cable, where a neighbourhood shares capacity. Between roughly 7pm and 11pm you may see meaningfully lower figures than at 3am, and that is your provider's capacity planning rather than a fault.
Worth measuring deliberately: if the difference is dramatic and persistent, it is evidence worth raising with them.
Two implementation details explain a lot of apparent weirdness.
A single connection rarely saturates a fast line. TCP takes time to ramp up, and one stream often cannot fill a gigabit connection at all. Tests open several parallel connections to compensate — which is realistic for a browser loading a page, and less so for a single large download.
Averaging hides the shape. A test reporting one number as an average over ten seconds tells you nothing about whether throughput was steady or wildly variable. Our speed test shows throughput as it is measured rather than only at the end, so a connection that starts fast and collapses looks different from one that is genuinely steady.
If wired results are consistently far below what you pay for, at several times of day, with nothing else running, you have something worth reporting. Take the readings with you when you call.
Why is my speed test lower than my advertised speed?
Advertised figures are usually "up to" and measured to the router under ideal conditions. Wi-Fi, device limits and overhead all reduce what you see. Within about 20% of the advertised figure over Ethernet is normal.
Which speed test is most accurate?
None is authoritative, because each measures the path to its own servers. Use the same test at the same time of day and watch the trend — consistency tells you more than any single number.
Should I test over Wi-Fi or Ethernet?
Ethernet to measure your connection. Wi-Fi to measure what you actually experience. Testing both and comparing tells you which one is your problem.
Why is my upload so much slower than my download?
Most consumer connections are deliberately asymmetric. Cable and DSL allocate far more capacity to download. Fibre is usually more balanced.
My speed is fine but video still buffers. Why?
Probably latency or bufferbloat rather than bandwidth. Check whether latency rises sharply while something is downloading; if it does, queue management on your router will help more than a faster package.