Third-party resources
Third-party resources are scripts, styles, fonts, images, and frames served from a domain other than the site’s own — analytics, tag managers, ads, consent banners, video embeds, chat widgets, A/B-testing and personalization tools. They are the part of page-weight the site author uses but does not author, and the answer to the wiki’s open question: a large, near-universal share of typical bloat is third-party.
The data
Per web-almanac-third-parties-2024 (HTTP Archive 2024):
- 92% of pages load ≥1 third party; popular sites carry more (median 66 for the top 1,000 vs. 27 for the top million).
- Third-party requests are mostly scripts (30.5%) and images (26.0%).
- Google owns 5 of the top-10 third-party domains (
googletagmanager.com,google-analytics.com,googleapis.com,google.com,youtube.com); Meta’sfacebook.comis the only other in the top five.
Why they resist minimalism
Third parties pull in more third parties: median inclusion-chain depth 3.4, with 14% of chains deeper than 5. So much third-party weight is indirect — never directly chosen by the author. This is why first-party tactics like hand-authoring less (landing-page-14kb) or smaller bundles (bundle-size, squint) don’t touch it: third-party weight is a separate governance problem (consent, tag managers, vendor contracts) with “privacy, security, and performance implications.”
A performance-budget can cap the number of third parties, but enforcing it means organizational discipline, not just a build flag. See synthesis for how this splits the minimalism thesis into a first-party layer and a third-party layer.