The Core Web Vitals Obsession Has Hit Diminishing Returns for Most Sites

I've watched engineering teams spend a full sprint shaving 200 milliseconds off an LCP that was already comfortably in the 'good' threshold, while a genuinely slow, unoptimized checkout flow sat untouched because it wasn't a Core Web Vitals metric. The sprint produced no measurable ranking change and no measurable conversion change, because the site had already crossed the threshold where Core Web Vitals meaningfully affects either.

The threshold effect everyone ignores

Google's own messaging on Core Web Vitals has always framed it as a threshold signal, not a continuous ranking factor to optimize indefinitely — crossing from 'poor' to 'needs improvement' or from 'needs improvement' to 'good' matters. Going from a good LCP of 2.3 seconds to an excellent 1.1 seconds mostly doesn't move rankings further, because you're optimizing a metric past the point where it's doing any work. The marginal engineering hour spent there is close to wasted relative to almost anything else that hour could have gone toward.

Crossing into 'good' on Core Web Vitals is worth real engineering effort. Polishing an already-good score is mostly effort spent for a graph that looks nice in a report.

Where that effort is usually better spent

• Fixing pages still sitting in 'poor' or 'needs improvement' — this is where almost all the real ranking and UX benefit still lives.
• Investing in actual conversion-path performance (checkout, signup, search), even when those flows score fine on Core Web Vitals but are slow or clunky in ways the metric doesn't capture.
• Content and internal linking work, which for most sites moves rankings more than another round of performance polishing once you're already in the 'good' band.
• Treating Core Web Vitals as a quarterly health check rather than a continuous optimization target once every page is comfortably in the good range.

I still take Core Web Vitals seriously — getting a site out of 'poor' is some of the highest-leverage SEO work there is. The mistake is treating a metric with a known threshold effect as something to optimize without limit, long after the marginal return has flattened to approximately zero.