How InternetSpeedGeo measures your connection
A result describes the connection between this browser and a Cloudflare measurement endpoint at the time of the test. It is a useful observation, with limits that matter when comparing connections.
Measurement engine and timing
New tests use the Cloudflare Speed Test JavaScript library. The interface allows approximately 30 seconds: separate windows of about 10 seconds for latency, download and upload. Each window has a finite set of requests; the requests can finish before the displayed phase ends. The timer does not mean data flows continuously for 30 seconds.
What the numbers mean
Download and upload use actual transferred data and the library’s throughput summaries, converted from bits per second to Mbps. Eight megabits equal one megabyte. Ping and jitter use the library’s latency summaries in milliseconds. Lower delay and less variation usually help interactive tasks. Loaded latency and packet loss are not measured by this configuration.
The separate ping tool
The ping page times 12 HTTP requests to this website, including response handling. It reports their mean, an upper middle sample as the median, and average absolute change between successive samples as jitter. This is not ICMP ping and its destination differs from the full speed test.
Country and provider comparisons
Provider rows show arithmetic averages of accepted submissions from the last 90 days, grouped by country, provider and the broadband or mobile mode chosen by visitors. Moving the map or searching a city does not create city-level measurements. Directory entries and APNIC population estimates are not speed-test samples.
Sample quality and missing values
Submissions need a recognized provider, plausible values and an accepted method; duplicate IDs and rapid repeats are limited. These checks are not independent certification of every browser submission. Older M-Lab results remain eligible alongside new Cloudflare results, so a historical aggregate can mix methods. No samples means no speed average; a small voluntary sample is not representative of all customers.
Making a useful comparison
Pause backups and other transfers, keep the device and connection mode consistent, and compare Ethernet with Wi-Fi if available. Record whether a VPN is active and repeat at different times. Browser load, radio conditions, routing and the measurement endpoint affect results. A test uses real data and may consume a substantial part of a metered allowance.
Corrections
Send a page URL, the disputed statement and supporting evidence to hello@internetspeedgeo.com. The review date records this page’s content check; it does not certify individual measurements.