Measured speeds, by country, with the sample size.

How these numbers are made

Every figure on this page is passive. Devices already carrying traffic report what that traffic achieved, and those readings are aggregated per country.

There is no active test that downloads a file to produce a number. That would spend a traveller's data on a measurement they did not ask for, and it would only describe one instant in one place.

Readings are grouped by country rather than by city, because a city median for anywhere outside a handful of capitals would rest on a few dozen samples. A number built on a few dozen samples is a rumour with a decimal point.

Where a country has enough readings for a city view, that appears on the individual coverage page rather than in this table.

The window is 90 days rolling. That is long enough to survive a bad week during a festival or a network upgrade, and short enough that a genuine improvement shows up within a quarter.

Countries with large sample bases move slowly. Countries with small ones move visibly, which is exactly why the sample column is not hidden.

Read the tenth percentile, not the headline

The median is the middle reading. Half the time you do better, half the time you do worse. It is a fair summary and a poor predictor of the moment you actually care about.

That moment is arriving at a busy airport, being on a train with four hundred other people, or standing in a stadium concourse. The tenth percentile answers it. It is the speed you get in the worst tenth of readings.

If it sits above 5 Mbps, video calls hold. If it sits above 2 Mbps, maps and messaging are comfortable and streaming is not. If it drops below 1 Mbps, expect the phone to feel broken even though the status bar looks full.

The practical rule is that anything above about 25 Mbps median is indistinguishable from anything else above 25 Mbps median for phone use. A 119 Mbps country and an 80 Mbps country feel identical in the hand.

Where they differ is in how far the tenth percentile has fallen away from the median. That gap is the number to compare.

Why nobody else publishes this

Publishing measured speed is uncomfortable, because it produces countries where your own service looks ordinary. Third party testing has put Airalo between 10 and 22 Mbps, and Holafly near 7 Mbps before throttling and under 1 Mbps after.

Neither company prints those figures anywhere on their own site. We print ours, because a traveller deciding between two products deserves a number rather than an adjective.

The countries where we look ordinary are the countries where we are working hardest, and hiding the figure would not make the connection any faster.

If you want the per country view with the local traps, city breakdowns and the carriers in priority order, every row in the table links to it.