The Core Web Vitals myth: when LCP fixes actually move rankings
Most performance work doesn't move traffic. The work that does has a very specific shape — and it's almost never the work the dashboards push you toward.
Every quarter a team I'm advising shows me a deck with a beautiful CrUX chart — LCP down from 3.1s to 2.4s, FID gone, CLS under 0.05 — and asks why organic traffic didn't move. The honest answer is: nothing in that chart implied it would.
Core Web Vitals are a ranking factor only in the sense that they're a tiebreaker. When two pages are otherwise close — same intent match, comparable authority, similar freshness — the faster one wins. The faster page wins more often if your competitors are slow. It does not win at all if your page wasn't already in striking distance.
When LCP work actually moves rankings
I keep a small spreadsheet of every CWV project I've reviewed where organic traffic clearly moved within eight weeks. There are three patterns:
- The page was on position 4–8 for a high-intent commercial query, and a single competitor on the same SERP was demonstrably faster. Closing that gap is the only one that consistently moves you up two spots.
- The site shipped a templated improvement across tens of thousands of pages at once — usually a server-side render path replacing client-side hydration. The gain shows up as a global lift in impressions, not as ranking changes on tracked queries.
- The Pagespeed regression was new (a recent deploy) and the ranking drop was new (the same week). Fixing the regression buys back the drop, no more.
Notice what's not on the list: improving LCP from 2.6s to 2.1s on a page that was already passing. That's the work most teams spend their quarter on, and it's almost always invisible in traffic.
How to localize an LCP regression to a deploy
When something does go wrong, the diagnosis loop should not take a sprint. Pull the CrUX p75 series for the URL pattern, find the inflection week, then list every deploy in that window. In our experience the culprit is almost always one of: a new third-party script, a font swap, a hero image that lost its priority hint, or a hydration boundary that moved.
The mistake I see most often is treating CWV as a hygiene project — a checklist that lives in the SEO team's quarterly plan. It's not hygiene; it's a control loop. Wire alerts to the p75 changes you care about, fix regressions in the same sprint they ship, and stop spending engineer-quarters on improving green numbers.