How we measure: where the numbers in our articles come from

Our articles are full of numbers, and nearly all of them are our own. This page answers the fair question a reader has: who wrote this, and why these numbers can be trusted.

Who writes this

Mayak is made by a small team. We set up the nodes in the Netherlands, Poland and Russia ourselves, answer support mail ourselves, and write these articles ourselves. We have no separate «content authors»: the write-up is done by whoever fixed the fault.

We do not put personal names on the site, and that is deliberate: the organisation answers for the text as a whole, and you can check us by the numbers rather than by a name.

Source one: our own connection log

The app reports how a connection attempt ended: which step worked, how many milliseconds each stage took, whether the network was mobile or Wi-Fi, and which app version it was. The shares of the form «so many attempts out of so many succeeded» come from that log.

What is not in that log and never will be: no site addresses, no traffic contents, nothing about who opened what. We do not collect it — details in the privacy policy.

The log has a finite depth, and we say what it is: it starts on 21 August 2026, and we will not promise «over 30 days» where two weeks of data exist.

Source two: a phone bench

Some conclusions cannot come from a log — «does moving to another port help», for instance. Those we check by hand: a phone, SIM cards of different carriers, the same connection profile in several variants, and a written record of what happened.

The largest such measurement was made in Moscow in late July 2026 on three networks. The write-ups on MegaFon, MTS and Beeline stand on it. One city, one phone, one week — and the articles say so plainly, because such a measurement cannot be stretched over a whole country.

Source three: live lines

The nodes are ours, and we check them with a live tunnel rather than with tests: bring the connection up, push traffic through it, look at what a person would see. That rule appeared after a day on which service checks reported «all good» for a node that was running the wrong binary.

The rules we keep

• A number carries a date and a window. «85 %» without a window is not a number.
• Next to a number stands what it does not say. Every article has a «what these numbers do not say» section, and it is not decoration.
• Somebody else's claim and our own measurement are different things. If a fact comes from another site, we write «they say» instead of passing it off as measured by us.
• We do not measure other people's services. We are not able to measure another company's app, so there are no «who is better» rankings here.
• Negative results get published too. When a check shows that a popular piece of advice does not work, we say so — even when staying quiet would be more convenient.
• Our own fault is named as ours. When part of the failures fell inside our own outages, we count them separately and say so in the text.

What we never measured

This list is shorter than the previous one, and it matters more: it shows the border of what we know.

• Tele2 (t2) — we have no SIM of theirs and never ran a single measurement; what follows from that is a separate write-up.
• Phone vendor shells (Xiaomi, Samsung, Honor and the rest) — we have no per-vendor measurements, only the overall picture from the log.
• Other people's apps and their servers — see the rules above.
• What the carriers changed and when — we cannot see their settings and will not pass guesses off as facts.