What a sustainable WordPress build weighs
We hear the same WordPress reputation repeated often: plugins pile up, page builders ship code for features nobody uses, and the resulting page sends several megabytes before a visitor has read a word. We think some of that reputation is earned. The question we wanted to answer about our own site was a narrower one, with an answer we could check rather than assert: what does a WordPress build weigh when it is built lean on purpose, and how does that compare to how WordPress sites typically ship?
This is a sustainable WordPress case study with one subject: agnikii.co.uk, our own homepage, measured on three dates in October 2026. We are not claiming our build is typical, because it is built specifically not to be. We are showing the bytes and being honest about the one part of the build that is not yet where it should be.
What agnikii.co.uk actually weighs
On 6 October 2026, we ran a Lighthouse audit of our own agnikii.co.uk homepage (mobile, simulated throttling, Lighthouse 12.8.2) and measured 475,821 bytes across 18 requests, with a performance score of 91. On 11 October, we ran the same homepage through Lighthouse repeatedly with the same settings, and 19 of 20 runs recorded between 471,827 and 471,981 bytes. Our own monthly measurement series separately records our homepage at 458 KB on 2 October, also at a performance score of 91, a weight change of -29% against our prior month’s reading.
None of these figures is rounded up or averaged across a sample. They are dated readings of one page, stated with their method, which is the standard this post holds itself to throughout. If you want the general case for thinking in bytes before anything else, see page weight in web design; this post applies that thinking to one specific WordPress build.
We are not breaking that total down by resource type here. The per-category figures sit in our working audit record rather than our published measurement series, and we only quote figures from the published series. What we can say is what the total does not contain: there is no page builder in the stack, so there is no page-builder runtime in those bytes. The general principle behind why total weight matters, lightweight web design sets out in more depth; this post stays with the one build.
What the build actually is, and is not
The theme behind agnikii.co.uk, `agnikii-digital-flex`, is a custom-built WordPress theme using Advanced Custom Fields (ACF) Flexible Content to assemble page sections, with no page builder plugin in the stack and a deliberately short plugin list. It is a classic PHP-template theme, not a WordPress block theme and not a full-site-editing build; the layouts are defined in template code and populated through ACF fields, not assembled from Gutenberg blocks and `theme.json`.
We are specific about this because the distinction matters to anyone evaluating the comparison: the weight advantage here comes from a lean, hand-built theme and a restrained plugin count, not from a particular WordPress editing paradigm.
That is also why we link rather than repeat our own custom web design vs WordPress themes piece here: it sets out when a theme is the right call and when custom work earns its cost, and we do not want to re-run that argument. This post assumes you already have, or are building, a custom WordPress theme, and answers a different question: what does a build like that actually weigh once it is live, measured rather than promised.
How that compares with WordPress generally
The HTTP Archive’s Web Almanac, in its CMS chapter, found WordPress sites carry a median mobile page weight of approximately 2,894 KB, which includes 638 KB of JavaScript and 119 KB of CSS, among the platforms it compared. Our own homepage, at under half that median, is not explained by anything special about WordPress itself; it is explained by what we chose to install, and not install, on top of it.
HTTP Archive data shows that WordPress performance variance is driven more by configuration than by core limitations.
We looked for a reliable figure for how many plugins the average WordPress site runs, and for how much weight a typical page builder adds, and could not find one we trusted. Every plugin-count figure we checked traced back to a round-up article repeating an earlier round-up article, never to a named survey or a disclosed sample.
The page-builder weight claims we checked were thinner still: the only one with any stated method behind it was a single, small, dated case study on a specific low-spec server, which is not a sound basis for a figure about page builders generally. Rather than repeat either kind of number with a caveat attached, we have left both out.
What the Web Almanac’s data does show is that images and scripts dominate the typical WordPress payload, so that is where a lean build starts. The coding discipline behind keeping that payload down generally, rather than WordPress specifically, is covered in sustainable coding practices and greener website code.
Where our own build is not yet doing what we’d recommend
We are not going to publish a sustainable WordPress post that only shows the flattering numbers. So here is the uncomfortable one: the Proxima Nova files behind our body text are self-hosted .woff files, not .woff2. Google’s current web.dev guidance on font performance is unambiguous on the format question: use WOFF2 and nothing else, because it compresses roughly 30% better than WOFF. Our headline font, Salo Paolo, is already WOFF2, so the fix here is converting the Proxima Nova files, not rethinking the approach.
There is a second font question worth being straight about, because it runs against a piece of advice repeated often enough that it has become received wisdom: that self-hosting fonts is simply the lightweight choice over loading them from a third-party provider.
Google’s own current guidance does not support that as an unqualified rule. It states plainly that “the performance differences between these two options are less clear-cut” than the advice suggests, and that self-hosting only outperforms third-party hosting where the site already has a CDN and HTTP/2 in place. Our fonts are self-hosted from the theme, which is a reasonable choice given our hosting setup, but we are not going to claim it as a universal best practice; it is not, because the source we would cite to support it does not say that.
We’d convert the Proxima Nova files to WOFF2, subset them to the characters we actually use, and limit the number of weights we load, which is exactly what web.dev’s own guidance recommends: “use only WOFF2 and forget about everything else.” That is the next change to this build, not a hypothetical one. It is the same discipline lean web design describes more generally: cut what the page does not need before reaching for anything more drastic.
What this means for carbon, and why we keep the two figures apart
The monthly measurement series models this homepage’s modelled carbon at 0.069 g CO₂e per view on grid electricity, or 0.057 g CO₂e per view with the renewable-hosting credit applied, since Green Web Foundation lists our host (20i) as green. We say modelled deliberately, every time, because that is what it is.
We measured the bytes above directly, with Lighthouse, on a stated date, with a stated tool version. The grams are not measured; they are produced by running those measured bytes through the Sustainable Web Design model, version 4, via the `@tgwf/co2` package, version 0.19.0, which makes assumptions about grid intensity, device energy and data centre efficiency that are the model’s, not a direct reading of anything. Reducing unnecessary bytes supports both numbers from the same lever: less data to send is both the faster page and the lower modelled footprint, not two separate benefits that happen to point the same way.
The takeaway
A WordPress build does not have to weigh what the ecosystem’s median implies. We think the Web Almanack’s own framing backs that up: weight follows configuration, not the CMS. Our own homepage sits at under half a megabyte with no page builder in the stack, built on a custom theme with ACF Flexible Content rather than a block or full-site-editing approach, and that is a specific, checkable choice rather than a property of WordPress itself. It is not a finished state.
The fonts prove that, named plainly rather than hidden. If you are weighing up whether your own WordPress site needs a lean rebuild or a lighter theme, custom web design vs WordPress themes is where that decision gets worked through; our sustainable web design service is where we do the measuring and the building.

