Three numbers, three different questions
What does a speed test actually measure?
A speed test measures how one device exchanges test data with one server, over one network path, at one moment. It is a useful snapshot, not a complete verdict on your internet connection.
Start here
The Short Version
- Ping is round-trip delay. Lower usually means a more responsive path to that test server.
- Download is the rate at which test data reaches your device. Upload measures the other direction.
- Mbps means millions of bits per second. It describes a data-transfer rate, not how quickly every app will feel.
- Server location, test design, Wi-Fi, VPN routing, competing traffic, and device load can all change the result.
Fastest useful move
Check This First
Run the same test twice under the same conditions. If the readings are far apart, one run is not enough to describe the connection. Repeat it before changing anything else, then look for traffic or conditions that changed between runs.
Try it here
Measure the browser-to-edge path
This site's tool checks latency, download, and upload between your browser and the Cloudflare edge serving the request. That makes it a focused path check—not a measurement of every destination on the internet.
Follow the test
What's Actually Happening
A browser speed test creates synthetic traffic—data made for the test, not your files or messages. It times that traffic as it moves between your device and a selected server.
- 1 Latency Times a round trip
- 2 Download Edge sends 8 MiB
- 3 Upload Browser sends 4 MiB
For download and upload, transferred bits divided by elapsed time gives Mbps.
Every part of that setup matters. A nearby server, a long test, or several simultaneous transfers can produce a different number from a distant server or one short browser transfer. Neither result has to be fraudulent; the tests may be asking different questions.
Change one thing at a time
The Troubleshooting Ladder
- Repeat the same test. Two or three runs reveal whether the first number was typical or a brief swing.
- Pause obvious competing traffic. Cloud backups, game downloads, video streams, and other users can share the same connection.
- Move closer to the Wi-Fi access point. If the result improves, the local wireless link is part of the problem.
- Compare Wi-Fi with Ethernet when possible. A wired run helps separate the internet service from the room's radio conditions.
- Compare VPN on and off, if policy allows. A VPN can change the route and introduce another endpoint that affects both latency and throughput.
- Try another device and another test. This separates one browser or device limit from a network-wide pattern and shows whether server choice matters.
- Record the conditions, not just the number. Keep the time, device, connection type, VPN state, test server, and repeated results before escalating the problem.
For the curious
Under the Hood
“Speed” bundles several different network properties into one casual word. Separating them is what turns a result into a diagnosis.
Bandwidth, throughput, and latency are not synonyms
Bandwidth or capacity
How much traffic a path can carry. The link with the least available capacity can become the bottleneck.
Throughput
The transfer rate actually achieved by this test, over this path, during this interval.
Latency
The time data takes to travel. A ping result normally reports a round trip in milliseconds.
Jitter and loss
Jitter is variation in delay; loss means some packets never arrive. Both can disrupt real-time traffic even when Mbps looks healthy.
What Mbps measures
Mbps means megabits per second. There are eight bits in
a byte, so it is not the same unit as MB/s, or megabytes
per second. A 100 MB file contains about 800 megabits. At a perfect,
steady 100 Mbps, the transfer would take at least eight seconds;
real transfers also have overhead and changing conditions.
What this site's test does—and does not do
The current tool takes the median of four browser round trips, downloads an 8 MiB payload, and uploads a 4 MiB payload. The three stages run in sequence against the site's Cloudflare edge path.
- Other websites and services
- Jitter and packet loss
- Latency while the connection is busy
It does not measure packet loss, jitter, latency while the connection is busy, your modem's line rate, or the performance of a particular game, meeting service, streaming provider, or website. Those require different measurements or different destinations.
Match the symptom to the signal
Common gotchas
When Things Get Weird
The speed test is great, but one site is slow
The site may use another route, server, or content provider. A healthy test path cannot clear every other destination.
Download is fine, but upload is weak
The directions are measured separately. Your service plan may list different download and upload rates, so check it before assuming a fault.
Calls lag on a fast connection
High throughput cannot cancel high delay, delay variation, or packet loss. Peak Mbps is only one part of real-time quality.
Results fall when someone starts a backup
The test and the backup are sharing capacity. Delay can rise too when queues fill, even before the connection feels “fully used.”
Every test service gives a different answer
Compare server location, route, transfer duration, and number of simultaneous streams. Different methods can produce different but valid snapshots.
Only one device tests slowly
Its radio, distance, browser, power mode, CPU load, or security software may be the bottleneck rather than the internet service.
Receipts
Source Notes
The interpretation above combines published measurement guidance with the current implementation of this site's browser test.
- The published speed-test methodology documents the current sample count, payload sizes, calculation, and sequence used by this site's browser test.
- The FCC's Measuring Broadband America technical appendix for HTTP download and upload test methodology, Mbps throughput, and the limit that a test measures only the path between its client and target.
- Measurement Lab's speed-test accuracy and methodology guide for the effects of server placement, network path, and single- versus multi-stream testing.
- RFC 9318 for the distinction between capacity, latency, working latency, and higher-order application metrics.
- RFC 9439 for end-to-end round-trip delay, delay variation, available bandwidth, and throughput terminology.
- Apple's Wi-Fi aggregate-throughput guidance for the effects of shared airtime, device connection rate, interference, distance, and other devices on observed throughput.