ISP Observatory · Methodology

Boringly honest, on purpose

Every number in the rankings can be traced to a measurement from a real home. Here is exactly how they're made — including the parts that don't flatter us.

Real homes, real links

Every data point comes from a small sensor node on someone's home connection — cable, fiber, DSL or LTE. No datacenters, no test servers, no lab conditions.

Continuous, not one-shot

Each node probes well-known anchors (1.1.1.1, 8.8.8.8, 9.9.9.9) every few minutes, around the clock. A daily aggregate is the median of thousands of samples — a bad evening can't hide and a lucky minute can't win.

The Wi-Fi hop is subtracted

Nodes also ping their own home router. That local hop is removed from the score, so a provider isn't punished for a customer's old router in a far corner of the flat.

Minimum sample, visible n

An ISP enters the ranking only with 5+ measured homes, and the node count is always shown. Below the threshold we display data as "collecting" — informative, but never ranked.

Known bias, stated

Node hosts are enthusiasts and may sit on better-than-average plans. Rankings compare providers within the same self-selected population — they're a strong signal, not a national census.

Open data pipeline

Measurements come from the open-firmware Sensmos sensor network. Aggregates update daily; provider identity is derived from the network (ASN), never self-declared.

The metric, precisely: for each home we take the 24-hour median round-trip to the anchors, subtract that home's own Wi-Fi hop, then take the median across all homes on the same provider (ASN) in a country. Jitter and packet loss are averaged the same way. History builds from daily aggregates, so a 7-day trend arrow appears once a provider has a week of data.
Want to see it in action? Pick a country and read the numbers — every ranked row shows its sample size.
View rankings →