Download
How fast data reaches you — web pages, video, file downloads. It is the number providers advertise because it is the one that is easiest to sell, and it is rarely the constraint in a business.
Four figures describe a connection, and most people only look at one of them. The headline download number is the least useful of the four for anything other than streaming.
How fast data reaches you — web pages, video, file downloads. It is the number providers advertise because it is the one that is easiest to sell, and it is rarely the constraint in a business.
How fast data leaves you — video calls, cloud backup, sending files, hosting anything. Most residential plans are heavily asymmetric, which is why a fast connection can still make you the frozen person on a call.
The round trip to the test node, in milliseconds. It governs how responsive everything feels. A 20 ms connection at 50 Mbps feels quicker to use than a 150 ms connection at 500 Mbps.
How much the ping varies between packets. It is the figure that decides whether voice and video hold up. Under 10 ms is good; above 30 ms is where calls start breaking up and people blame the software.
Bandwidth requirements are smaller than most people expect, and the constraint is usually upload or latency rather than download.
| Activity | Rough requirement | What actually matters |
|---|---|---|
| Email, browsing, SaaS apps | 5–10 Mbps per person | Latency matters more than speed here — a 20 ms connection feels faster than a 200 Mbps one at 150 ms. |
| HD video calls | 3–4 Mbps up and down per call | Upload is the constraint. Jitter above ~30 ms causes the robotic audio people blame on the platform. |
| 4K streaming | 25 Mbps per stream | Download only, and buffering absorbs short dips — streaming is the least latency-sensitive thing you do. |
| Cloud backup, large file transfer | As much upload as you can get | Most residential plans are heavily asymmetric, which is why a 500 Mbps plan can still take all night to back up. |
| VoIP phone calls | 100 kbps per concurrent call | Almost no bandwidth, but it is the most sensitive thing on the network to jitter and packet loss. |
| Office of 20 people | 200–500 Mbps, with upload ≥ 100 | Concurrency, not headline speed, is what runs out. Measure at 3pm, not at 7am. |
Wi-Fi is the bottleneck far more often than the circuit. A wired test is the difference between a provider ticket that goes somewhere and one that gets closed as no fault found.
One result is a snapshot. Run it morning, afternoon and evening across a few days — the pattern tells you whether you have a fault, congestion, or a connection that was never what you thought.
A backup, a sync client or someone streaming will show up in your result. If you cannot quiet the network, note what was running — it is part of the measurement.
Results from this tool stay in your browser, so you can show a provider a pattern over weeks rather than one bad afternoon. That is what moves a support case.
The most common cause is Wi-Fi rather than the line itself. A browser test measures this device, over this connection, at this moment — so the wireless link, the age of the device, and whatever else is using the network all count. Test again on a wired connection before concluding the circuit is at fault. Provider speeds are also quoted as 'up to', and are measured to their own equipment rather than across the internet.
Because they measure to different places. A result is always the speed between you and one specific test node, so the route, that node's capacity and how busy it is all affect the figure. Tests also report different statistics — some show a peak, some a 90th percentile, some an average. Comparing the same test over time tells you far more than comparing two different tests once.
Ping is the round trip to the test node. Under 30 ms is good, under 50 ms is fine for everything except competitive gaming, and over 100 ms is noticeable in conversation. Jitter is how much that round trip varies between packets. For voice and video, jitter matters more than raw speed — under 10 ms is good, and above 30 ms is where calls start breaking up.
More than most plans provide. Video calls, cloud backup, VoIP, remote desktop, sending large files and anything hosted on your own premises all consume upload. An office running video calls and cloud backup wants at least 100 Mbps up, and symmetric connections are worth paying for once more than a handful of people depend on them.
Your test history is stored in your own browser so you can compare runs over time, and one button clears it. The large transfers go directly between your browser and a public measurement network, not through intSignal, so we never see your traffic. One lookup is made to a third-party service to identify your network operator. Nothing about your test is sent to us.
Shared infrastructure. Cable and fixed-wireless connections in particular share capacity with neighbours, so evening speeds can be a fraction of the morning figure. If a connection is fast at 7am and slow at 8pm, that is congestion rather than a fault, and it is worth documenting with repeated tests before raising it with a provider.
We design, procure and manage business connectivity — circuits, SD-WAN, failover and the QoS that keeps voice usable when the line is busy. If your results are persistently below what you pay for, or calls break up on a connection that tests fine, that is a diagnosis worth doing properly.