Skip to content

Engineering

The cost of a slow page, measured properly

Performance is usually discussed as a lab score. What matters is the experience of a real user on a mid-range phone and an ordinary connection.

Hyperzen Innovation2 Oct 20261 min read
The cost of a slow page, measured properly

Performance work often stalls because it is framed as a score to raise rather than an experience to fix. The score is a proxy. The experience is the thing.

Measure what users get

Lab tools run on fast hardware and stable networks. Real users are on mid-range phones, congested mobile networks and browsers with a dozen tabs open. Field data — Core Web Vitals from actual sessions — tells you what is happening; lab data tells you why.

Where the time usually goes

  • JavaScript. Shipped, parsed and executed before anything is interactive. Most of it is framework and dependency weight, not application code.
  • Images. Correct formats and explicit dimensions eliminate both weight and layout shift.
  • Fonts. A blocking font request delays every word on the page.
  • Third-party scripts. Tag managers and widgets frequently cost more than the entire application.

A practical order of work

  1. Render on the server and ship interactivity only where it is needed.
  2. Set a JavaScript budget and enforce it in CI, so regressions are blocked rather than discovered.
  3. Serve modern image formats with correct sizing and dimensions.
  4. Self-host fonts, preload the critical face, and always define a fallback.
  5. Audit third-party scripts twice a year and remove what nobody defends.

None of this is novel. It is simply rarely done consistently, and consistency is what separates a fast site from one that was fast at launch.

Found this useful? Share it.

//Start a conversation

Have a problem worth solving?

We would rather talk about your system than write about ours.

Enter ↵ to send