In mid-October, Opensignal published a study comparing 4G and 5G fixed wireless access (FWA) with Starlink across 14 European markets. Starlink was the fastest rural option in all 14, at 82 Mbps, about twice the speed of the average FWA provider, as reported by ISPreview. But the leading FWA providers beat Starlink on a second measure, Consistent Quality, in the urban and rural areas of most markets.
That split is the useful part. A connection can be fast and still be uneven, and a slower one can be steadier. Which one a user prefers depends on what they run.
What the study measured
Opensignal uses crowdsourced data from users of its app and scores two things: download speed and Consistent Quality. Opensignal describes the second as showing "how often it is good enough during everyday use." ISPreview's report does not give the score threshold, the sample sizes or the study period, so the percentages below should be read as Opensignal's own scale, not a universal one.
The study covers only FWA delivered over mobile networks (in the UK, Vodafone, Three UK and EE). It excludes smaller dedicated FWA networks, some of which can deliver gigabit speeds. Starlink's results are therefore compared with mobile-network FWA, not with all fixed wireless.
The UK numbers
The UK figures show how the two measures can diverge. As reported by ISPreview:
- Three UK had the fastest urban FWA download at 81.2 Mbps and the fastest rural FWA download at 32.4 Mbps.
- Starlink measured 80.8 Mbps.
- On Consistent Quality, Three UK led urban areas at 68.8% and EE led rural areas at 62.1%. Starlink scored 68.1%.
In cities, then, FWA and Starlink were close on both measures. In the countryside, Starlink was far ahead on speed, and its Consistent Quality score was above the best rural FWA score. ISPreview reports that rural FWA is limited by spectrum and network density. Across all 14 markets, though, the leading FWA providers usually scored higher on consistency than Starlink.
Why an average hides the problem
A speed test reports one number from one moment. Real-time applications care about every moment. A video call is carried in a steady stream of small packets, and the call degrades when delay or loss rises, even briefly. A file download can tolerate that; a conversation cannot.
The scenario described here is illustrative. Two links each average 40 Mbps over a test. Link A ranges between 38 and 42 Mbps. Link B swings between 8 and 72 Mbps.
Link A: 38 ÷ 42 = 0.90
Link B: 8 ÷ 72 = 0.11
Both links would post the same average, and a one-shot speed test could rank either one first. Only one of them is predictable. The Consistency of Service measure used here (minimum divided by maximum over a test) is not the same as Opensignal's Consistent Quality score, which is built from crowdsourced user experience. They point at the same idea from different directions.
Speed also depends on delay
Throughput is not set by line speed alone. A single TCP session can move only as much data per round trip as its window allows, so round-trip time (RTT) caps its rate. With a 64 KB window (65,535 bytes, or 524,280 bits) and a 25 ms RTT:
That ceiling is the same on a 100 Mbps line and a 1 Gbps line. If RTT also varies from moment to moment, as it can on a congested or shared access link, the ceiling moves with it. This is why latency and its variation matter as much as the headline rate. Neither the Opensignal study nor ISPreview's report gives latency figures, so this article does not compare the two technologies on delay.
What to measure before choosing a rural link
If you are connecting a remote site, a branch or a home worker over FWA or satellite, the study suggests asking more than "how fast". Measure over time, at the hours your users are actually online, and record the lowest result as well as the average.
MyConnection Server sends its own test traffic between a test point at the site and an endpoint you choose, and reports RTT, jitter, packet loss and equilibrium throughput, the sustained rate the path can hold, rather than a one-second burst. Its bandwidth quality test also reports Consistency of Service. Because tests can be scheduled every few minutes or hourly, a software satellite at the site can build the time series that a single speed test cannot, and thresholds can raise an alert when results fall outside what you expect.
Opensignal's data shows what its app users experienced on average. It cannot show what your site experiences at 9 a.m. on a Monday. That needs your own measurements.
Sources: study findings are as reported by ISPreview, October 2026, from Opensignal's FWA vs Starlink in Europe analysis. Accurate as of October 26, 2026.