Why two speed tests disagree
Different results can both be real. A result depends on endpoint, route, browser, traffic and measurement method; the largest number is not automatically the most accurate.
What to check
Do not run tests simultaneously. Keep device, network and VPN state fixed, then run three tests with short gaps. Record server names and times. A different endpoint can use different routing or peering. Browser extensions, CPU load, Wi-Fi airtime and the number of parallel transfers can also change throughput.
Worked example — illustrative, not a measured result
Results of 92, 95 and 180 Mbps do not show a reliable 180 Mbps baseline. The median is 95 and the wide spread deserves investigation. Conversely 92, 94 and 95 is a more repeatable observation. These are hypothetical numbers, not provider rankings or promised tolerances.
What to do next
Compare like with like and keep outliers in your notes rather than hiding them. Repeat during busy and quiet periods. Avoid combining mobile and fixed-network tests in one conclusion. Local report averages summarize your stored tests; country averages use voluntary submissions and may mix methods, so they cannot establish the speed at your street or address.