It works. But does it load fast?

Slow and janky pages lose users and rank worse, and most of the causes are visible in source long before you measure them. The Frontend Performance scanner reads your code for the patterns that hurt load time and smoothness, before you deploy, offline, on every scan. It is the source-side complement to the Pro Live Site Auditor, which measures your real Core Web Vitals on the deployed URL.

What it catches

Nine static checks for the patterns that slow a page down or make it jump around:

Render-blocking scripts CSS @import Layout shift (no img dims) No font-display Full-library imports key={index} No lazy-loading Large inline base64 Many stylesheets

The ones that cost you the most

1. Render-blocking scripts in the head

A <script src> in <head> without defer or async stops the browser from parsing the page until it downloads and runs. Every synchronous head script pushes back the moment the user sees anything. The fix is usually one attribute.

2. Images with no reserved space

An <img> with no width, height, or aspect-ratio has no space reserved, so the page jumps as each image loads. That is cumulative layout shift (CLS), one of Google's Core Web Vitals, and it feels janky even when it isn't measured.

3. Bundle bloat from full-library imports

import _ from 'lodash' or import moment from 'moment' pulls the entire library into your bundle with no tree-shaking, often hundreds of kilobytes shipped for a few functions. On a mobile network that is real load time. Deep imports (lodash/debounce) or lighter libraries fix it.

The honest line

This is a readiness check, not a benchmark. Static analysis catches the patterns that commonly hurt performance; it does not measure your actual load time. The real numbers (LCP, CLS, INP on a real connection) come from the Pro Live Site Auditor's Core Web Vitals. These findings are quality signals: they sharpen your grade but never gate the Ship-It security verdict.

Built for precision, not nagging

Only high-confidence patterns. It deliberately skips the false-positive-prone advice ("you didn't useMemo", "virtualize this list") that erodes trust. If it flags something, it's real.
Count-gated where it should be. A single dimensionless logo is fine; it flags a page full of them. A hero image stays eager; it flags an image-heavy page with no lazy-loading anywhere.
Free, offline, every scan. No URL, no build step. Pairs with the Pro Live Auditor, which measures the outcome on your deployed site.

How to fix what it finds

Add defer/async to head scripts (or type="module"), and replace CSS @import with a <link>.
Set width and height (or aspect-ratio) on content images, and add loading="lazy" to below-the-fold ones.
Add font-display: swap to your @font-face rules so text shows immediately.
Import only what you use (deep imports or tree-shakeable packages), and use stable list keys instead of the array index.

IOnclad reports patterns it can see in source. It is not a profiler and does not measure real-world performance.

Ship something that loads fast.

Scan your code in under a minute, offline. Nothing leaves your machine.

Try the Browser Scanner → Download for Windows