Core Web Vitals in 2026: which one actually moves the needle
Google's March 2026 update bumped INP to a ranking factor and tightened the thresholds. Everyone's scrambling. Below is the order we'd prioritize if you only have time to fix one or two — based on what we see across the work we ship.
The order we'd fix things
If you can only chase one metric this quarter, chase these in this order:
1. INP (Interaction to Next Paint) — the replacement for FID. INP under 100ms feels instant. Above 300ms, the page feels broken even if it isn't. JavaScript that blocks the main thread is the usual culprit.
2. LCP (Largest Contentful Paint) — the one everyone talks about. Sub-1s LCP is the new floor for "feels fast". The jump from 2.5s to 1.5s usually matters more than the jump from 1.5s to 0.8s.
3. CLS (Cumulative Layout Shift) — less directly tied to conversion than we expected, but strongly tied to bounce on mobile when content jumps under a thumb.
What's harder than it looks
INP is hard because it's driven by JavaScript execution. Modern frameworks ship a lot of JS upfront, and if any of that blocks the main thread on first interaction, you're in trouble. The three biggest wins we see in practice:
- Move analytics and marketing tags to
<script async>or a tag manager. - Break up long tasks with
scheduler.postTaskor explicit yielding. - Ship less JavaScript, especially to mobile.
What's easier than it looks
LCP is mostly about the hero image. Preload it, serve AVIF, size it correctly, don't hide it behind a client-side query. That alone gets most sites under 1.5s.
CLS is almost always about images/embeds without explicit dimensions. Set width and height on everything. Done.