Skip to content
IT

Is my connection stable?

A speed test lasts five seconds, and in five seconds a problem that shows up occasionally does not show up at all. This page watches the connection for two minutes, one measurement a second, and shows you the pattern. It is the test to run when meetings freeze but the speed test says everything is fine.

In plain words

Response time (latency, often called ping): how long a packet of data takes to travel to the server and back. Stability (jitter): how much that time changes from one measurement to the next. Missed reply: a request that gets no answer within this test's waiting time. It can come from the network, from a delay that is too long, or from a momentary hiccup at the server, and it does not automatically amount to a lost packet: saying that needs measurements at a lower level, which a browser cannot reach.

Typical time—ms
Variation—ms
Worst—ms
Missed replies—

Pattern

The chart appears as soon as measurements start.

How to read the result

The typical time is the middle value of the measurements: half are faster, half slower. Under 40 milliseconds is good, between 40 and 100 is normal for a home line, above 150 video calls and games become hard work. I use the middle value rather than the average because one terrible measurement moves the average and does not move the middle, and the middle describes better what the connection is usually like.

The variation is the number that matters most for talking to someone. It measures how much response time changes from one second to the next. A line at a steady 20 milliseconds is excellent for a meeting; a line swinging between 20 and 200 is worse than a slow but steady one, because the person listening hears your voice arrive in lurches.

Missed replies are the requests that got no answer in time. Zero out of a hundred and twenty is what you expect on a healthy line. A few happen. Many, especially all together at one point of the chart, say the connection had a hole in it, and that is exactly the moment the video call froze.

In the chart, the bars are the individual measurements and the height is the response time. Pale bars mark missed replies. The shape matters more than the numbers: a regular row is fine, isolated spikes are fine if they are rare, a series of rhythmic spikes or a step that never comes back down are the two pictures worth attention.

What this test does not see

This test measures the road between your device and the nearest Cloudflare node. If the problem lies beyond that, say between your provider and the video call platform's servers, it will not show up here. That is not a flaw in this page: from a browser you cannot measure a road you do not travel.

And it measures at rest, that is, while you are not doing anything else. The most common trouble with video calls at home is latency under load: response time going through the roof exactly when somebody is downloading or uploading something big. To see that you need the trick below.

The useful trick:

Start this test and, while it runs, put the connection under strain: start a big upload to a cloud service, or ask somebody at home to watch a video in high definition. If the chart jumps up and stays there, you have just seen with your own eyes why meetings freeze, and the cause is inside your house.

And if the numbers are bad

If the variation is high only over Wi-Fi and disappears with a cable, the problem is the Wi-Fi, and the checks in order are in Slow Wi-Fi: six checks to make before calling your provider. If it is high over cable too, and gets worse under load, the full reading of what is happening and what you can ask your provider are in Video calls freeze even though the line is fast.