TTFB Test

TTFB varies from one visit to the next: whether the request hit a warm cache, whether the server was busy with another visitor, how the route happened to run. Two consecutive measurements can differ by a factor of two.

That is why this test makes five requests in a row and shows the median, with all individual measurements and the spread next to it. When the fast and slow requests differ a lot, the tool points it out: a server like that serves some requests from cache and computes others from scratch.

Where exactly the time goes

A long TTFB does not mean the server is at fault. Before it even sees the request, the browser goes through a DNS lookup, a connection setup, and a TLS handshake. The tool splits the total time across these four parts and names the largest one.

That changes the conclusion entirely. If the time goes to the server response, look at the database, caching, and heavy queries. If it goes to the connection, look at the network channel, the certificate, and the name servers; the site itself has nothing to do with it.

What counts as a good value

• under 800 ms: good, the Google threshold for comfortable loading;
• 800 to 1800 ms: tolerable, but noticeable;
• over 1800 ms: slow, the visitor sees a blank screen.

TTFB sets the lower bound for everything else: until the first byte arrives, the browser has nothing to render. Improving page speed without sorting out TTFB first usually gets you nowhere.

Measuring from another country

If your site has visitors abroad, it helps to know what response time they see. The test repeats the measurement through a remote location; that path to the location is part of the result, so the number comes out higher, and you should compare it against the same kind of measurement for another site rather than against your local one.

Next to this tool there is a website speed test that looks at the whole page, images and scripts included.