Third-party domains
How many distinct third-party domains the page contacts.
At a glance the headline numbers for Third-party domains
How many distinct third-party domains the page contacts.
The typical site contacts 7 third-party domains. The heaviest 1% contact 36 or more.
Distribution & median INP site count and median INP at each level of third-party domains
Passing INP by third-party domains which level passes the INP most often
Third-party domains 7. p75 11. p99 36. Spearman with INP r = +0.30. At the low end (1): INP 43ms. At the high end (>p98): INP 74ms. computed
Why this matters for the Core Web Vitals, and where to start fixing it
Third-party domains are a common Core Web Vitals issue. First of all, each third-party domain needs a new connection. That means time is lost on a DNS lookup, a TCP handshake and a TLS negotiation. If you had served those resources from your main domain, the browser could have re-used the connection that is already open (that is why you want HTTP/3: one fast, re-usable connection for everything you control). When that domain serves a font or a stylesheet, the browser cannot render before the connection is ready. The connection setup delays the LCP.
Also, third-party resources are usually out of your control. They can become a SPOF (single point of failure) when their server is slow or down. They can also change in size or behaviour without your knowledge, and your Core Web Vitals change with them. Self-hosting fixes both problems at once. Preconnect only hides part of the connection cost.
How do third-party domains affect the Core Web Vitals?
Passing LCP barely moves across the range: 89% at one end, 87% at the other. This signal does not separate passing sites from failing ones.
The effect is bigger on CLS. With a single third-party domain, 86% of sites pass it. Among the sites with the most third parties, 73% do. Third parties inject ads, embeds and consent banners, and injected content moves the layout after the page looks done.
Chrome field data from 94,910 sites, representing millions of real page loads. How we measured.