Site speed gets sold as a ranking factor, which undersells it. As a ranking signal it is a tiebreaker. As a conversion factor it is one of the larger levers available on most sites.
The metrics that carry commercial weight
Largest Contentful Paint is when the main content appears, and it correlates most directly with abandonment. Interaction to Next Paint measures whether the page responds when someone taps, which matters disproportionately on mobile. Cumulative Layout Shift measures whether content jumps under a finger, which produces mis-taps and lost trust at exactly the wrong moment.
Where the returns stop
Going from six seconds to two is transformative. Going from 1.2 seconds to 0.9 is usually invisible to a human being and is an engineering exercise rather than a commercial one. Knowing where you sit on that curve determines whether performance work is a priority or a distraction.
The usual culprits
- Unoptimised hero images shipped at desktop resolution to phones
- Third-party tags: chat widgets, heatmaps and duplicate analytics
- Web fonts loaded without a display strategy, blocking the first render
- Render-blocking scripts in the document head
- Client-side rendering on pages that could be static
How to decide
Measure field data rather than lab scores. A perfect Lighthouse result on a fast laptop tells you very little about a customer on a phone in a regional area. If your real-world LCP is above 2.5 seconds on mobile, performance work will likely pay for itself. If it is under 1.5, spend the money on something else.