Performance work gets filed under engineering housekeeping, which is why it is so often deferred. But every second of delay is a share of visitors who never see the page, never read the offer and never had the chance to convert.
The framing that gets it prioritised is simple: speed is not a technical score. It is the percentage of your paid and organic traffic that survives the wait.
The three numbers, in plain language
Core Web Vitals sound abstract until you translate them into what the visitor experiences.
- LCP — how long until the main thing on the screen appears. Aim under 2.5 seconds. This is usually your hero image or headline.
- INP — how long the page takes to react when someone taps or clicks. Aim under 200 milliseconds. This is what 'laggy' actually means.
- CLS — how much the layout jumps around while loading. Aim under 0.1. This is why people tap the wrong button.
Measure them on real visits, not only in a lab tool on a fast laptop. The gap between a developer's machine and a mid-range phone on mobile data is where most performance problems live.
The fixes that actually move the numbers
Most sites do not need exotic optimisation. Four categories account for the vast majority of real-world slowness.
- Images. Serve modern formats, size them to the space they occupy, set explicit width and height, lazy-load everything below the fold and never lazy-load the hero.
- Fonts. Limit families and weights, self-host, preload the one used above the fold, and use a fallback display so text is readable while the font loads.
- Third-party scripts. Chat widgets, tag managers, heatmaps and pixels are often the single largest cost on the page. Audit them quarterly and delete the ones nobody reads.
- JavaScript. Ship less of it, split it by route, and render the content that matters on the server so it does not wait for a bundle.
The hero image deserves its own paragraph
On most marketing pages, LCP is the hero image, and the hero image is the one thing teams optimise last because it is the one thing they want to look perfect.
The compromise is not to use a worse image. It is to serve the right size to each device with responsive sources, compress it properly, give it priority loading, and reserve its space in the layout so nothing shifts when it lands.
Speed is a maintenance problem
Sites do not become slow at launch. They become slow over eighteen months, one tracking script and one uncompressed banner at a time, each added by someone with a good reason.
Set a budget and enforce it: a maximum page weight, a maximum number of third-party scripts, and a check in the release process. A budget makes the trade-off explicit at the moment someone wants to add something, which is the only moment it can be argued about honestly.
What to do this week
Run a real-device test on your top three pages over a throttled mobile connection. Note the LCP element, the largest three assets and the full list of third-party domains being contacted.
In most cases that single page of notes contains a week of work with a bigger conversion impact than a redesign — and unlike a redesign, you can prove the result.