The WP Rocket Settings That Actually Move Core Web Vitals

A caching plugin has thirty settings and roughly six of them change a Core Web Vitals score. Knowing which six, and which two are dangerous, is most of the work.

Measure first

Record your starting numbers on a representative page — not the home page, which is usually atypical. You need LCP, CLS and INP from field data where you have it, and lab data otherwise. Without a baseline you cannot tell which change helped.

The six that matter

Setting	Affects	Risk
Page caching	TTFB, LCP	None. The core feature.
Delay JavaScript execution	INP, LCP	High. Test thoroughly.
Remove unused CSS	LCP, render blocking	Medium. Can break layouts.
Preload LCP image	LCP	None. Always enable.
Font optimisation	LCP, CLS	Low. Host fonts locally too.
Lazy loading below the fold	LCP, bandwidth	Low, if you exclude the LCP image.

1. Page caching

The default and the whole point. Verify it is actually serving cached pages rather than just being switched on — check for the plugin’s cache header in the response, or compare TTFB logged-out versus logged-in.

2. Delay JavaScript execution

Defers scripts until the first user interaction. This is usually the largest single improvement to INP, and it is the setting most likely to break something.

What breaks: sliders that must animate on load, anything that renders content with JavaScript, some analytics, and occasionally the cookie banner. The fix is an exclusion list, built by testing rather than guessing.

3. Remove unused CSS

Generates critical CSS per page type and loads the rest asynchronously. Good for LCP, and it sometimes strips rules that are applied conditionally — hover states, elements revealed by JavaScript, admin-bar styles.

Check the pages with the most conditional interface after enabling it, not just the home page.

4. Preload the LCP image

Tells the browser to fetch the largest above-the-fold image immediately rather than discovering it during parsing. Free improvement, no downside, and on image-led pages it is often the single biggest LCP gain.

Make sure that image is also excluded from lazy loading. Lazy-loading the LCP image is one of the most common self-inflicted LCP problems.

5. Fonts

Host fonts locally rather than loading from a third-party CDN: one fewer connection to establish, and no third-party dependency. Set font-display: swap so text renders in a fallback immediately, and preload the one or two faces used above the fold.

6. Lazy loading

Everything below the fold, nothing above it. Set explicit width and height on every image regardless, because that is what prevents layout shift.

What does not move the needle

• Database cleanup. Useful housekeeping, irrelevant to Core Web Vitals.
• Minification, mostly. Compression at the server already does the heavy lifting.
• Emoji script removal. Real, and measured in a handful of kilobytes.
• Heartbeat control. Reduces server load, not visitor-facing metrics.

The unglamorous truth about CLS

CLS is rarely fixed by a caching plugin. It is fixed by reserving space: explicit image dimensions, fixed heights on ad containers, and avoiding content injected above existing content after load. Advertising is the usual culprit on content sites.

Common questions

Will this work with another caching plugin?

The settings have different names, the principles are identical. Page cache, defer JavaScript, critical CSS, preload LCP.

Do I need a CDN as well?

It helps TTFB for distant visitors. If your audience is concentrated near your server, the gain is modest.

Why did my score get worse?

Usually critical CSS generated against a page that has since changed, or a cached version predating a fix. Regenerate and purge, then re-measure.