How is internet speed measured?

Your plan says "up to 100 Mbps"; the speed test shows 62. Your neighbour has the same package and gets 85. Where do these numbers come from, and what does the test actually do behind the scenes?

What does Mbps mean?

Speed tests report megabits per second (Mbps): millions of bits moved per second. The common confusion: file sizes are written in bytes (B), and one byte is 8 bits. So a 100 Mbps line downloads at most 12.5 megabytes of file data per second under ideal conditions. Your browser showing "12 MB/s" while the speed test says "96 Mbps" is not a contradiction — they are two units for the same thing.

What the test does

A modern speed test runs three measurements in sequence:

1. Ping (latency): Tiny requests are sent to the server and the round-trip time is recorded in milliseconds. The first few requests carry setup costs — DNS resolution, the TLS handshake — so they are excluded from the average.
2. Download: Large blocks of data are requested and the bytes arriving per unit of time are counted. A single connection usually cannot fill the line — packet loss and latency limit how fast one TCP stream can go. That is why tests open 4–6 parallel connections; Ookla and fast.com do the same.
3. Upload: The same logic in reverse — the browser sends randomly generated data to the server. The data is deliberately random: if a proxy along the way could compress it, the result would look better than reality.

The warm-up allowance: TCP slow start

When a TCP connection opens it does not know the line's capacity; it starts cautiously and grows its sending window each round trip. This is called slow start. The first second or two of a test is spent on this ramp, so good tests exclude that window from the average — otherwise short tests would underestimate the line.

Why results vary

Even two back-to-back tests on the same line can differ. The main reasons:

• Wi-Fi: Walls, distance and interference from neighbouring networks cut wireless throughput substantially. The most accurate measurement is over an Ethernet cable.
• Other traffic on your network: A test run while another device streams 4K video measures the remaining bandwidth, not the whole line.
• Peak hours: Regional congestion rises in the evening.
• Test server location: A distant or busy server tells you about the path, not your line. The test on this site measures against one of the world's most widely distributed networks (Cloudflare); your traffic is routed to the nearest point.

A short checklist for reliable results

• Use a cable if possible; otherwise move closer to the router.
• Pause downloads/updates and turn off your VPN.
• Repeat the test at different hours and look at the average.
• Read the result in Mbps; divide by 8 when comparing with MB/s.

A single test is a photograph; regular tests are a film. Only the trend over time shows your connection's true character. Tests you run on this site are stored anonymously on your device — the history section shows the trend.