What is website latency? (and what counts as good)
Latency is the round-trip time between a visitor and a server — distinct from bandwidth. Here's what adds to it, typical values, and how to reduce it.
The definition
Latency is the time a message takes to travel from a visitor to a server and back — the round-trip time. It's a measure of delay, not capacity: bandwidth is how much data a connection can move per second, latency is how long each exchange takes. A connection can have huge bandwidth and still feel slow if every round trip takes 300ms.
It's also distinct from TTFB. TTFB includes latency plus the time the server spends generating a response; latency is just the travel time. A fast server behind a slow network path still delivers a poor TTFB.
What adds latency
Physical distance is the floor: signals in fiber travel at roughly two-thirds the speed of light, which works out to roughly 1ms of round-trip time per 100km of cable path. No optimization removes that — only moving the endpoints closer together does.
On top of distance, real connections stack round trips: a DNS lookup, a TCP handshake (one round trip), and a TLS handshake (one more round trip with TLS 1.3, two with older versions). The last mile adds its own share — Wi-Fi interference and cellular networks typically contribute more variable delay than the wired backbone.
What counts as good
As rough bands for a single round trip: under about 20ms is typical within the same region, 40-80ms across a continent or the Atlantic, 100-150ms across the Pacific, and 250ms or more to the opposite side of the globe. There's no single "good" number — a value that's excellent from Sydney to Frankfurt would signal a problem from Paris to Frankfurt.
A full HTTPS connection costs several round trips before the first request is even sent, so the latency a visitor experiences on a fresh connection is a multiple of the raw network round-trip time. That's why shaving even 30ms of distance shows up noticeably in page loads.
How to reduce it
Distance is the biggest lever: hosting near the audience, or putting a CDN in front so requests terminate at an edge location near the visitor, cuts the round trip itself. The rest is spending fewer round trips — reusing connections with keep-alive instead of re-doing handshakes, TLS 1.3's shorter handshake, and HTTP/2 or HTTP/3 multiplexing many requests over one connection.
Common questions
- Is latency the same as ping?
- Related but not identical. Ping measures raw network round-trip time with ICMP packets. Web-focused latency tools typically measure TCP and TLS connection round trips over HTTPS, which reflects what browsers actually experience — usually slightly higher than an ICMP ping to the same server.
- What is a good latency for a website?
- It depends on distance. Under roughly 20ms is typical within the same region, 40-80ms across a continent, and 250ms or more to the opposite side of the globe. Judge a measurement against where it was taken from, not against a universal number.
- Does more bandwidth reduce latency?
- No. Bandwidth is capacity, latency is delay — upgrading a connection moves large files faster but doesn't shorten round trips. Interactive browsing is usually limited by latency, not bandwidth.
Measure round-trip latency to your server.
Measure your latency →