Interpret response time in PageSpeed, Pingdom and GTmetrix

Reviewed on September 8, 2026 2 min read tiempo de respuesta rendimiento

TTFB measures the time until the first byte of a response. Depending on the tool and the point from which it starts counting, it may include DNS, connection, TLS, redirects, network and server work. A phase called Waiting in a network waterfall may exclude stages that are shown separately. Check the metric's definition before comparing figures.

In applications such as Magento, WordPress or PrestaShop, a large part may come from PHP, SQL queries, plugins, external calls or a lack of caching. TTFB alone does not let you conclude that the server is slow or that the application is the only cause: you need to repeat the measurement from several locations and check it against logs and internal metrics.

A ping measures the round-trip time of ICMP packets to the destination when that protocol is allowed. It is useful for assessing latency and packet loss, but it does not measure the time taken by PHP, MySQL, TLS or the website's HTTP delivery. There is no universal threshold: a reasonable latency depends on the distance, the network route and the test location.

Prepare a useful comparison

Use the same final URL, session, device and location. Note whether the response comes from the browser cache, a CDN or the server, and repeat several times: a first visit without cache is not equivalent to a response that is already cached.

In the network waterfall, separate DNS, connection, TLS, waiting and download. Compare a public page with another operation that runs PHP. A difference points the diagnosis in a direction, but on its own it does not show which component needs to change.

If the waiting time is in PHP, use X-Ray and check the resources for the same interval. If the HTML arrives quickly and the page still does not respond well, check images, CSS and JavaScript in the browser.

When you ask for help, include the URL, the time, the test region, the exact metric and several samples. Do not compare a local ping with a load test from another continent as if they measured the same thing.