Heavy Lifting: The A–Z of Everything Quietly Adding Weight to Your Website
Photo: Texas. Office of the State Auditor; Alwin, Lawrence F, Public domain, via Wikimedia Commons
Page weight creeps. It never arrives in one go. A theme update here, a marketing tool there, a video someone in sales insisted on — and eighteen months later your mobile LCP is sitting at 4.8 seconds and nobody can remember why.
The frustrating part is that most of the culprits are things you added deliberately and would defend in a meeting. Below is the full alphabet, followed by the honest bit: which of these actually matter.
The A–Z
A is for autoplay video. A five-second looping hero clip is usually 3–8MB. It also competes with your text for bandwidth, which is exactly the wrong priority.
B is for background images straight off the camera. A phone photo at 4032px wide, dropped into a hero slot 1400px wide. Compression would have saved 90% of it and nobody would have noticed.
C is for carousels. They load slides two and three whether or not anyone sees them, and the analytics almost always show people don't reach slide two.
D is for drag-and-drop builders. Convenient, but many ship the entire component library on every page, used or not.
E is for extra webfonts. Each family and weight is a separate download. Four weights across two families is a common, and mostly unnecessary, tax.
F is for form and CAPTCHA scripts. Some load on every page rather than only the contact page where they're needed.
G is for tag sprawl. Years of 'just add this pixel' in Google Tag Manager, each one pulling in its own library.
H is for hidden content that loads anyway. Tab panels, accordions and mobile-only menus that fetch their images regardless.
I is for icon fonts. A whole font file to render fourteen small glyphs you could have inlined.
J is for JavaScript chat widgets. Often the single heaviest third-party script on a small business site, and often used twice a week.
K is for keeping things switched on. Plugins and features nobody has touched since 2023, still loading on every visit.
L is for lazy-loading the hero image. Lazy loading is good — except above the fold, where it delays the exact thing LCP is measuring.
M is for map embeds. A full Google Maps iframe on the contact page, when a static image and a directions link would do the job.
N is for nested redirects. Not weight, but delay. Three hops to get to a page is three round trips on a 4G connection.
O is for omitting image dimensions. No width and height attributes means the layout jumps as images land — a Core Web Vitals issue in its own right.
P is for pop-up libraries. A full modal framework to show a newsletter box to people who arrived four seconds ago.
Q is for query strings that defeat caching. Assets served with a changing query parameter get re-downloaded by browsers and CDNs alike.
R is for review widgets. Embedding a third-party review carousel often drags in a framework you didn't ask for.
S is for social embeds. An embedded Instagram post brings its own scripts, styles and tracking along for the ride.
T is for tracking pixels. Each one is cheap alone. Eight of them is a different conversation.
U is for unused CSS. A framework's full stylesheet on a page that uses a tenth of it.
V is for vendor scripts loaded synchronously. Anything in the head without async or defer blocks the page from rendering at all.
W is for wrong formats. Photographs saved as PNG. Logos saved as enormous PNGs with transparency nobody needed.
X is for excess DOM nodes. Deeply nested page-builder markup can push a simple page past 2,000 elements, which slows rendering on cheaper phones.
Y is for YouTube embeds using the full player. There's a well-known lighter pattern that loads a thumbnail first and the player on click.
Z is for zombie plugins. Deactivated but still in the folder, or active with no purpose anyone can name.
Which of these actually matter
Here's the uncomfortable news: not all of the alphabet is equally guilty. Core Web Vitals care about three things — how fast the main content appears (LCP), how much the page jolts about (CLS), and how quickly it responds to a tap (INP).
The heavy hitters for UK SMEs, in rough order of impact:
- The hero image or video — dominates LCP on mobile
- Synchronous and third-party scripts — delay everything and hurt INP
- Images without dimensions — the usual cause of layout shift
- Font loading — late-swapping fonts cause both jolt and delay
- Everything else — real, but usually a rounding error next to the four above
So before you spend a fortnight auditing icon fonts, open PageSpeed Insights on your three most important pages and look at LCP, CLS and INP on mobile. That tells you which letter of the alphabet is actually your problem.
A realistic fix order for a busy owner
Week one. Compress and resize the hero images on your top five pages. Convert photographs to WebP. Add width and height attributes. This alone often moves LCP by a second or more.
Week two. Open Tag Manager and count what's firing. Anything with no owner and no measurable purpose comes out. Ask each remaining vendor whether they can load on interaction instead of page load.
Week three. Trim fonts to the weights you actually use, and self-host them if they're coming from a third-party CDN.
Week four. Deal with the chat widget. Load it on click, or move it to a link, or lose it. Test both for a month and see if enquiries change.
Later. Carousels, pop-ups, review widgets, map embeds. Do these when you're next touching those pages anyway — they're worth fixing but rarely worth a standalone project.
The habit that prevents all of this
None of the above is hard. The problem is that weight is invisible until someone measures it, and nobody measures it until something feels slow.
So pick a date — quarterly is plenty — and record your homepage's total page weight and mobile LCP in a shared document. Two numbers. Next quarter, compare. When someone proposes a new widget, you'll have a baseline to argue with, and that's worth more than any audit.