Website Design in Vanderbilt, TX

The fallback font and the webfont have different metrics. The browser renders text immediately in a system font, then swaps in the custom one when it arrives — and if the two have different character widths and line heights, every line rewraps and everything below moves, which registers as layout shift. The modern fix is tuning the fallback with size-adjust, ascent-override, and descent-override in a local @font-face rule so its metrics approximate the real font closely enough that the swap is nearly invisible. Preloading the font file shortens the window in which any of this matters.

Almost certainly, and it's usually the HTML being cached longer than intended. Aggressive cache rules applied wholesale — at the CDN edge, in a hosting-level page cache, or in a plugin — can serve a stale document for hours after the source changed. The right split is caching HTML briefly or not at all while caching static assets like CSS, JS, and images for a long time, with a version string or content hash in their filenames so a new deploy produces new URLs. That way updates appear immediately and the assets that never change stay cached. A cache purge on deploy is the belt-and-braces addition.

The MX records almost certainly went with the DNS. When a host takes over a domain's nameservers, the domain's entire record set gets rebuilt from that host's defaults, and unless the mail records were carried across deliberately, mail starts routing to a server that has no mailboxes on it — silently, from the sender's perspective. Website hosting and email are separate services that happen to share a domain name, and the DNS zone is where that separation lives. Before any nameserver change, export the current zone in full and confirm the MX records, plus any SPF and DKIM entries, are recreated on the new side.

Possibly not the image at all. LCP is whatever the largest element painted in the viewport turns out to be, and if the hero has a large headline set in a custom font over it, that text block can be the LCP element — in which case LCP is gated on the font downloading, not the image. The other frequent surprise is the difference between a lab test on a fast connection and real field data from actual visitors, which is what gets reported. Identify the actual LCP element in the browser's performance panel before optimizing; people routinely spend a week compressing the wrong thing.

It's counterproductive. The intent is usually to stop a wrong saved number being filled in, but the effect is forcing everyone to type their number by hand — on a phone, in a hurry, which is precisely when errors happen and precisely when people abandon. Browsers also increasingly ignore autocomplete="off" on ordinary contact fields because it was so often misused. The better setup is autocomplete="tel" with type="tel", which brings up the numeric keypad and lets the browser offer the saved number that people can edit if it's wrong. Accuracy comes from clear labels and a confirmation step, not from making entry harder.

It matters practically. Every family means more font files, more requests, and more bytes before text renders confidently, and four families with a couple of weights each is a real load cost for something visitors never consciously register. It also erodes hierarchy: when several typefaces compete, nothing reads as clearly more important, and hierarchy is what lets someone scan a page instead of reading it. Two families is plenty — one for headings, one for body — and one family used across a range of weights and sizes is frequently better still, and lighter. Restraint here buys both clarity and speed.