A speed test result consists of four separate measurements, and treating them as a single score overlooks most of what determines day-to-day performance. A connection with excellent download speed can still perform poorly on a video call, while a connection that is average on paper can feel smooth for gaming. The four measurements combine differently depending on the activity.
The Four Measurements
| Metric | Meaning | Matters Most For |
|---|---|---|
| Download | How fast data arrives. | Streaming and large downloads. |
| Upload | How fast data is sent outward. This is significantly lower than download on many home internet plans, particularly over 4G/5G or satellite broadband such as Starlink. Some broadband services provide matching upload and download speeds, while others offer higher download than upload. Leased lines consistently provide matching upload and download speeds. | Video calls, cloud backups, sending large files. |
| Latency (ping) | The round-trip time to a server, in milliseconds. | Real-time applications: games, calls, remote desktops. |
| Jitter | How much latency varies from one moment to the next. A connection with low average latency but high jitter can still feel unstable, because packets arrive at inconsistent intervals rather than a steady stream. | Same as latency — real-time, interactive traffic. |
Kbps, Mbps, and Gbps: What the Units Mean
Download and upload are measured in bits per second, and results scale between three prefixes depending on magnitude — Kbps (kilobits), Mbps (megabits), and Gbps (gigabits) — each one 1,000 times the one before it. A result shown as 850 Mbps and one shown as 0.85 Gbps are the same speed, just scaled differently so the number stays easy to read at a glance.
| Unit | Typical Of | Real-World Example |
|---|---|---|
| Kbps (kilobits per second) | Audio calls, smart-home sensors, basic IoT devices | A voice call over WhatsApp or Skype uses roughly 100 Kbps in each direction — a small fraction of a single Mbps. |
| Mbps (megabits per second) | Web browsing, SD/HD/4K video, video calls, most home broadband plans | Netflix recommends 3 Mbps for standard definition and about 25 Mbps for 4K. Most home broadband plans (50–500 Mbps) live entirely in this range — 1 Mbps = 1,000 Kbps. |
| Gbps (gigabits per second) | Fiber home plans, office LANs, large file transfers, data centers | A 1 Gbps fiber connection downloads a 20 GB game in under 3 minutes, versus roughly 45 minutes on a 100 Mbps plan — 1 Gbps = 1,000 Mbps. |
One easy mix-up: bits and bytes are not the same thing, and file sizes are usually shown in the latter. A file manager or download dialog reports MB (megabytes), not Mbps (megabits per second) — since a byte is 8 bits, a 100 Mbps connection downloads at roughly 12.5 MB/s, not 100 MB/s.
Why the Same Connection Rates Differently by Activity
Each activity has a different sensitivity profile. Streaming video depends almost entirely on sustained download bandwidth and is largely unaffected by 100ms of latency. A competitive game depends heavily on latency and jitter and requires minimal bandwidth — a few hundred kilobits per second is generally sufficient. A video call requires both: enough upload bandwidth to send video, and low, stable latency so audio does not cut out or lag. This is why a single Mbps figure is a poor indicator of connection quality without knowing what the connection is being used for.
General Guidelines by Activity
| Activity | Download | Latency | Notes |
|---|---|---|---|
| Browsing | A few Mbps | Under ~100 ms | Keeps pages feeling responsive. |
| Streaming (HD) | ~5–10 Mbps sustained | Not critical | Buffering absorbs latency almost entirely. |
| Streaming (4K) | 25 Mbps+ sustained | Not critical | Recommended minimum for smooth 4K playback. |
| Gaming | Minimal | Under ~50 ms, jitter under ~10 ms | This is where a connection starts to feel genuinely responsive. |
| Audio calls | Under 1 Mbps, each direction | Low | Highly sensitive to jitter and packet loss — stability matters more than raw speed. |
| Video calls | 1.5–4 Mbps, roughly symmetric | Low | Asymmetric connections (fast download, weak upload) are most noticeable here. |
Why a Longer Test Produces a More Accurate Result
Connections do not reach full speed instantly — TCP ramps up gradually over the first second or two, a behavior known as slow start, and a very short test can end while still ramping up, understating actual throughput. A longer measurement window, with the initial ramp-up period excluded from the average, produces a steadier, more representative result rather than one skewed by the connection's speed during its first fraction of a second.
Why Results Vary Between Runs
Local Wi-Fi congestion, other devices actively using bandwidth on the same network, and general internet congestion at different times of day are all normal sources of variation. A single test represents only a snapshot at that moment. If a result appears unexpectedly low, run the test again, ideally with other heavy network usage paused.
If download and upload appear normal but a connection still feels unresponsive, latency and jitter — not raw throughput — are usually the cause. A ping test against the specific service in question can help isolate the issue further.
Frequently Asked Questions
Why is my upload speed so much lower than download?
That's normal on most home internet plans, especially over 4G/5G or satellite broadband — providers allocate more capacity to download since most traffic (streaming, browsing, downloads) flows that direction. Leased lines are the exception and provide matching speeds in both directions.
My download speed is high but video calls still feel bad — why?
Download throughput and call quality depend on different things. A video call needs enough upload bandwidth plus low, stable latency — a fast download number doesn't fix a weak upload or high jitter, which is usually what actually causes calls to lag or cut out.
Why do I get a different result every time I run the test?
Local Wi-Fi congestion, other devices actively using the connection, and general internet congestion at different times of day all cause normal variation. A single result is a snapshot, not a fixed number — if one looks unexpectedly low, try again with other heavy usage paused.
Why does a longer test give a more accurate number than a quick one?
A connection doesn't reach full speed instantly — TCP ramps up gradually over the first second or two (slow start). A very short test can end while still ramping up and understate the connection's real throughput; a longer window averages that out.
What does jitter actually mean, and why does it matter?
Jitter is how much latency varies from one packet to the next, not the average latency itself. A connection can have decent average latency but high jitter, which still feels unstable — packets arriving at inconsistent intervals is what causes stuttering calls and laggy games, sometimes more than the raw latency number would suggest.
Run the Internet Speed Test to measure download, upload, latency, jitter, and activity-based ratings in one pass.