Hi there!
I’m having an issue with pages built dynamically based on HubDB. They are highly optimized, and when I test them locally, the desktop LCP is always below 1.5s. However, when I run a test in PSI, the first result always shows an LCP of 3–3.5s, while every following test is below 1.5s. I don’t have these problems on other pages that are not dynamically generated from HubDB.
Has anyone dealt with a similar issue and knows how to solve it? Google Search Console marks these URLs as needing improvement because the LCP is above 2.7s.
Hey @AWebStudio,
Thanks for posting in the Community!
Tagging in a few experts here who might have some insight for us!
@GiantFocal, @Stephanie-OG, and @jonchim - any ideas that might explain this behavior?
Shane, Senior Community Moderator
Hi, @AWebStudio
Since the repeat PSI tests are already dropping below 1.5s, the issue is probably more related to the very first page load rather than the overall optimization itself.
Since it is only happening on HubDB pages, the cold request is likely hitting a server render delay before anything reaches the browser , worth checking your TTFB on that first hit specifically. Static pages skip this step which is probably why they are fine.
I would start with the basics first, preload the main hero image, make sure nothing above the fold is being lazy-loaded, and push as many non-critical scripts out of the initial load as possible. Usually the issue is that too many assets are competing during the initial render, which delays LCP. The browser just needs to stop tripping over itself before the first paint, that’s usually what kills LCP. The difference between the first PSI run and the repeat runs is often caused by how the initial request gets processed and loaded, so improving the first visible render usually helps a lot with Search Console as well.