How fast is your site from where your customers are?
Enter an address and we’ll measure it from five countries using real machines on real networks — then show you which part of the wait is distance and which part is your server.
What these numbers are, and what they aren’t.
Measuring a website badly is easy, and a confident wrong number is worse than no number at all. So here is exactly what we do and where it stops being reliable.
A median, never an average
Three probes run in each country and we report the middle one. Probes are volunteer machines on ordinary connections: in testing, three reported 29 ms while three others on the same continent reported 78, 84 and 267 ms. One bad probe moves an average and can’t move a median.
DNS is left out on purpose
Every probe uses its own resolver, and the same lookup varied from 2 ms to 285 ms in our own tests. That number describes the probe’s network, not your server, so including it in a headline would be dressing up noise as a finding.
Distance and server time, separated
The connect step is one round trip along the same path, so subtracting it from the time to first byte leaves roughly what your server spent working. It’s an estimate, not an instrument reading — but it separates the part nobody can fix from the part anyone can.
One moment, five points
This is a single sample from five countries at one instant, on a cold connection with nothing cached. It is not a verdict on your host, it won’t catch a problem that only appears under load, and it says nothing about how long the whole page takes to render.
Measurements run on the Globalping network, the same one we used for the numbers in our post on whether server location matters.
Want a second opinion on what you just saw?
Send us the result and we’ll tell you what we’d look at first — page cache, PHP version, images, database, or genuinely nothing at all. You don’t have to be a customer, and we won’t pretend a number is worse than it is to sell you something.