You click a link and there's a beat — a tiny pause — before anything happens. That pause is latency, and it's one of the most important (and most misunderstood) numbers on the internet. It's why a site can feel sluggish even on a fast connection, and why "just add more bandwidth" often doesn't help.
What latency is
Latency is the time it takes for data to travel from one point to another and back — the delay between a request and the start of a response. It's usually measured in milliseconds (ms), and lower is better.
A helpful frame: bandwidth is how much data you can move; latency is how long it takes to start moving. A wide pipe with high latency still feels slow. (More on that in bandwidth vs latency.)
What causes latency
Latency is the sum of several delays along the way:
| Cause | What's happening |
|---|---|
| Distance | Data physically travels at finite speed — far = slower |
| Network hops | Each router along the path adds a little |
| Server processing | Time for the server to think (TTFB) |
| Congestion | Busy networks queue traffic |
| Packet loss | Lost packets trigger slow retransmissions |
The big, unavoidable one is distance — there's a hard physical floor on how fast a signal crosses an ocean. That's exactly why serving content closer to users matters so much.
How to measure it
The classic tool is ping, which sends a packet and times the round trip (RTT). For the full picture across the route, traceroute and MTR show the latency at each hop, so you can see where the delay builds up.
And remember to look at percentiles, not averages — your slowest requests (p95/p99) are where users actually feel the pain.
How to reduce it
You can't beat physics, but you can shorten the trip and the wait:
- Use a CDN. Serving content from a nearby edge is the single biggest latency win for a spread-out audience — it shrinks the distance.
- Lower your TTFB. A faster backend (caching, optimised queries) cuts the server-think portion.
- Reduce round trips. Fewer requests, connection reuse, and modern protocols (HTTP/3) mean fewer back-and-forths.
- Put compute near users. Edge rendering and regional deployments cut distance for dynamic content too.
- Fix the network basics. Address packet loss and congestion that inflate effective latency.
The mental model: latency = distance + hops + server-think. To cut it, shorten the distance (CDN/edge) and shorten the think (faster backend). Those two moves cover most of it.
Why it matters
Latency compounds. Every extra round trip and every slow hop adds up, and users feel it as a site that "drags." It affects conversions, bounce rate, and the perceived quality of everything from page loads to API calls to video calls. Low latency is a feature.
The bottom line
| In one line | |
|---|---|
| What | The delay between request and response (in ms). |
| Biggest cause | Distance — data travels at a finite speed. |
| Measure | Ping (RTT), traceroute/MTR; look at p95/p99. |
| Reduce | CDN/edge (distance) + faster backend (TTFB). |
Latency is the quiet tax on everything online. You can't repeal physics, but you can serve content closer and respond faster — and that's most of the battle.
Measure it with the free ping and traceroute tools; go deeper in latency percentiles.