When you test your connection, beyond download speed you may have seen terms like latency, jitter and packet loss. They measure not how fast your network is but how stable it is. For gaming, video meetings and voice calls, these three often determine the actual experience more than download speed does.
This article explains what each one is, what counts as normal, and why a web-based test always reads higher than the command line.
Latency: how long a round trip takes
Latency is the time a packet takes to travel from your device to a target server and back, measured in milliseconds (ms). It measures responsiveness, not throughput.
A common confusion: latency and bandwidth are different things. Bandwidth is how wide the pipe is (how much data per second); latency is how long a signal takes to make the trip. You can have high bandwidth downloading large files quickly while latency is high and games stutter — one does not substitute for the other.
Physical distance is the main driver of latency. The speed of data through fibre has a ceiling, so intercontinental access is inherently slower than reaching a server in the same city. Accessing a US server from China takes over a hundred milliseconds in physical round trip alone — a hard floor no network can push below.
Rough reference ranges:
- Under 30 ms: excellent, same-city or nearby nodes
- 30–80 ms: good, imperceptible in everyday use
- 80–150 ms: normal, typical for cross-region access
- 150–300 ms: on the high side, noticeable in real-time applications
- Above 300 ms: clear lag, affecting video calls and gaming
Jitter: how steady the latency is
Jitter is how much latency fluctuates. If ten consecutive tests bounce between 40 ms and 120 ms, jitter is high; if each holds steady around 45 ms, jitter is low.
For real-time applications, jitter often matters more than latency itself. The reason: a steady high latency can be compensated by buffering — a video call caches a few tens of milliseconds of data and you never notice — but erratic jitter empties and fills the buffer unpredictably, producing choppy audio, stuttering video and "teleporting" in games.
An analogy: steady high latency is a speed-limited but flowing road you can anticipate; high jitter is a road that is fast one moment and jammed the next, never knowing what the next second holds. The latter is harder to live with.
So when reading latency figures, do not look only at the average — look at how steady it is.
Packet loss: whether data goes missing
Packet loss is how much of the data you send fails to reach its destination, usually given as a percentage.
In network transmission, data is cut into small packets sent separately. Ideally each arrives, but congestion, interference or overloaded equipment can lose some in transit. Lost packets must be retransmitted, and retransmission means delay and stutter.
Reference standards:
- 0%: ideal
- Under 1%: acceptable for most applications
- 1–3%: real-time applications (games, calls) start to be clearly affected
- Above 3%: noticeably problematic
A little packet loss is actually common on wireless networks; wired connections are usually steadier. If you have severe loss over Wi-Fi, plugging in an Ethernet cable often helps immediately.
Why a web test reads higher than the command line
This puzzles many people: testing the same latency, a web tool's result often runs tens of milliseconds higher than the command-line ping. It is not that the web tool is inaccurate — the two measure fundamentally different things.
The command-line ping uses the ICMP protocol, sending a minimal probe packet directly to measure the pure network round trip.
But browsers cannot send ICMP packets, for security reasons. So every web "ping tool" actually measures HTTP round-trip time — running the full sequence: DNS resolution, establishing a TCP connection, a TLS encryption handshake, and only then the actual request and response. That whole process running tens of milliseconds above a bare ICMP round trip is normal.
So:
- Command-line
pingmeasures pure latency at the network layer - A web tool measures the full round trip at the application layer, closer to how long it actually takes to open a website
Both numbers are meaningful; they just mean different things. A higher web-measured value does not mean your network is bad — it reflects how long it really takes to reach a service, which in a sense is closer to your actual experience.
Why some targets show "unreachable"
In a web ping tool, a target showing unreachable usually does not mean the service is down — it means the service is blocked on your current network.
That is actually useful diagnostic information. For example, if Google and YouTube show unreachable while other targets respond normally, you are on a network that restricts those services, rather than the services having a problem.
Conversely, if every target is unreachable, you can safely conclude the issue is your local network, firewall or proxy rather than a simultaneous outage across all of them — the odds of Google and your home router failing at once are far lower than the router alone having trouble.
So the combination of "which respond and which do not" itself helps you locate where the network problem lies.
How to test
Open the ping test and it measures round-trip latency from your connection to a set of major services (Google, TikTok, OpenAI, GitHub and more), shows jitter, and flags unreachable targets. It all runs in the browser with nothing to install.
A few tips:
- Test several times for stability. A single result is subject to momentary fluctuation; several consecutive runs reflect reality better, especially when you want to see jitter.
- Compare different targets. If only cross-border targets are slow while local ones are fast, it is an international-route issue; if everything is slow, it may be your local network.
- Compare across networks. Testing on Wi-Fi versus Ethernet, at home versus the office, helps you locate the bottleneck.
Summary
Latency, jitter and packet loss measure a network's stability rather than its speed: latency is round-trip time, jitter is the fluctuation in latency, and packet loss is the rate of dropped data. For real-time applications, jitter and packet loss often affect experience more than latency itself.
A web tool measures the full HTTP round trip including DNS, TCP and TLS, so reading higher than a command-line ICMP ping is normal — and closer to real access experience. For a quick look at your connectivity and latency to major services, use the ping test.