Why Speed Test Results Differ From Your Plan
It is common to run three speed tests and get three different answers, none of which matches the speed on your bill. That does not necessarily mean any of them is wrong. They are measuring slightly different things.
Why tests disagree with your plan
- “Up to” speeds. Providers advertise the maximum speed of a plan, measured under ideal conditions. Real-world results are usually somewhat lower, especially at busy times.
- Protocol overhead. Every packet carries addressing and control information alongside your data. On a typical connection this takes a few percent of the capacity, so even a perfect line measures slightly below its headline speed.
- Wi-Fi and devices. Your plan describes the line coming into your home. The test measures all the way to your device, including Wi-Fi, which is often the real limit. See how to improve Wi-Fi.
- Other traffic. Anything else using the connection during the test reduces the result.
Why different speed tests disagree with each other
They use different servers
Every speed test measures the path between your device and its server.
- Some tests use servers inside your provider’s network. These measure your line with very few other networks in the way, and often show the highest numbers.
- SpeedCheckWeb uses the nearest Cloudflare data centre, part of a large global network that many websites and apps also use. The result reflects a path similar to one used by real online services, which can include links between your provider and other networks.
- Some tests let you choose a server far away. Distance raises latency and can lower throughput.
They use different methods
Tests differ in:
- Duration: short tests may finish before a fast connection reaches full speed.
- Parallel connections: using several connections at once usually reaches higher numbers on fast lines.
- How the final number is calculated: an average, a peak, or a high percentile of short measurements. SpeedCheckWeb uses the average after a warm-up period, which tends to be more conservative than “peak” methods.
- Latency method: ICMP ping, small web requests, or measurements inside an open connection, each timed in different ways.
Conditions change between tests
Even back-to-back tests see different conditions: other devices waking up, Wi-Fi interference, or congestion somewhere along the route.
How to compare fairly
- Use a cable for any test you will rely on.
- Compare like with like: the same test, the same device, the same location, at similar times of day.
- Run several tests and look at the typical result, not the best one.
- Look beyond download speed. Upload, latency and latency under load often explain “slowness” better than download does.
- Read each test’s methodology. Ours is published in full on the methodology page.
So which result is “right”?
Each one is a genuine measurement of a particular path, at a particular moment, using a particular method. A test inside your provider’s network is best for checking your line. A test to a large content network like Cloudflare’s is closer to your everyday experience of websites and apps. If both are well below your plan on a wired connection, show both to your provider.
Sources
Figures from third parties can change; we review guides regularly. See our editorial policy.