How much does this site actually ship?

Page weight is one of the clearest indicators of how a site behaves on a weak mobile connection, and it is easy to measure without guesswork. When I shipped v2.0 of my studio site (theronradley.vercel.app), I wrote a build script to count every first-party request each route makes, sum the exact bytes on disk, and run gzip -9 over every asset. None of the numbers below are estimates.

STUDIO HOME PAGE · 167.3 KB RAW -> 80.6 KB GZIP (21 FIRST-PARTY REQUESTS) ■ Fonts 45.9 KB (27.4%) ■ CSS 40.3 KB (24.1%) ■ HTML 31.2 KB (18.6%) ■ JS 27.3 KB (16.3%) ■ Vector SVG 22.6 KB (13.5%) · Text pages: 8.2 KB JS Method: production build parser + level-9 gzip · zero third-party analytics
Cold-cache transfer budget: self-hosted fonts and CSS make up most of the fixed cost before any page content arrives.

Key findings


Route-by-route totals

Every page shares the same 45.9 KB font files and 40.3 KB stylesheet, so differences across routes come down to HTML length, interactive scripts, and SVG artwork:

RouteReqHTMLJSRawGzip
/2131.2 KB27.3 KB167.3 KB80.6 KB
/about/525.4 KB8.2 KB119.8 KB63.3 KB
/faq/543.0 KB8.2 KB137.4 KB65.0 KB
/research/518.4 KB8.2 KB112.8 KB61.7 KB
/research/site-weight/526.9 KB8.2 KB121.3 KB63.7 KB
/services/923.1 KB8.2 KB188.0 KB84.0 KB
/work/715.6 KB27.3 KB129.0 KB67.3 KB

Method and limits

After npm run build, a script parses each generated HTML file, resolves linked stylesheets, scripts, preloaded fonts, and SVGs, and writes raw plus gzip -9 byte totals to JSON in the repo. Live CDN responses use Brotli, which compresses a bit smaller, so gzip gives a conservative upper bound.

Transfer size also has clear limits: a build-time byte count does not measure server tail latency (which is what Wiremark tests), browser layout time, or warm-cache navigation. Still, keeping exact byte budgets in version control is one of the practical reasons I stick to hand-coded websites over page builders.


Linked mentions

You might also enjoy