The number you are shown
Ookla publishes every Speedtest result aggregated into ~600 m tiles, quarterly, as open data. We take the hotel's coordinates, find its tile, and report the test-weighted average download, upload and latency for that block.
If the exact tile is thin on tests we widen to the neighbouring tiles and tell you the radius we had to use. If nobody has ever tested there, we say so and give the nearest real measurement instead of printing a zero.
What this is not
This is not the hotel's WiFi. It is the broadband available at that address — the ceiling the hotel's connection sits under. What you get in the room is that number minus everything the hotel does to it: one circuit split across 200 rooms, per-device caps, a captive portal, ageing access points, and the 9 p.m. peak.
So treat green as this hotel could be good and red as this hotel cannot be good. Red is the more reliable signal: a hotel cannot deliver what the street does not have.
The verdict
Set by the worst of the three numbers, not the average. A 300↓/4↑ connection is a great download and a failed client call; averaging would hide exactly the failure you care about.
| Verdict | Download | Upload | Latency | Means |
|---|---|---|---|---|
| Call-ready | 100+ | 25+ | ≤40 ms | Calls, screen share and big pushes at once. |
| Workable | 40+ | 10+ | ≤70 ms | One video call is comfortable. |
| Risky | 15+ | 3+ | ≤120 ms | Audio survives; video drops at peak. |
| Tethering territory | — | — | — | Plan on mobile data as primary. |
Scores shrink toward the middle when a tile has few tests, so a single lucky test cannot produce a confident 95.
Real speedtests beat all of it
When someone submits an actual speedtest from inside a hotel, that replaces the neighbourhood estimate entirely and the page is marked ✓ measured. It is the only ground truth for hotel WiFi that exists anywhere, which is why the submit box is on every page.
Outages
Cloudflare Radar publishes outage annotations per country and ISP with start and end times. A national or ISP-wide outage takes the hotel down with it, so recent history is worth seeing. A hotel's own router dying will never appear there — no dataset tracks that, and we will not pretend otherwise.
Freshness
Ookla publishes quarterly and we keep 8 quarters, which is what lets a page show whether a location is improving or rotting. A weekly job polls for a new quarter, so one goes live within seven days of publication without anyone remembering to look. Places change; a two-year-old measurement is a rumour, not a fact.