Performance

PulseKeeper keeps two kinds of speed apart on purpose, because they answer different questions.

Server response time: every check#

How long your server takes to answer, measured on every check from every location. On the Insights tab you see it over time with the median, p95 and p99 (the average hides exactly the slow requests people complain about), and per location.

Page load speed: Google PageSpeed#

Once or a few times a day, depending on your plan (and once after a deployment, if you track deploys), Google PageSpeed loads the page like a phone would and reports:

• the performance score and Core Web Vitals: LCP, CLS, TBT and FCP;
• page weight: everything the browser downloads, split into images, scripts, fonts, CSS and HTML, and how it changed since the previous run;
• real-visitor data from Google, when the site has enough traffic for Google to report it;
• requests: every file the page loads, grouped by domain (this site or a third party), and a waterfall showing when each one started and finished, its type, status and size.

The waterfall timings come from Google loading the page without network throttling, while the score and LCP are simulated for a slow phone, so the two do not add up. Addresses are stored without their query string. We keep the waterfall for the last 10 runs; the score and page weight stay for the whole history.

PageSpeed measures from one place, not from our locations, so its numbers are never mixed with response time.

It does not run for a password-protected site: Google cannot send a password, so it would only measure the login prompt.

Changes after a deploy#

If you connect your repository, every deploy is measured before and after. See deploy tracking.