There have been questions around the rationale and data behind some of the decisions made in the Largest Contentful Paint metric. Here are some examples and data to help answer the questions.
- Why are body background images invalidated as candidates for LCP? Examples of pages with/without body background invalidated, showing problems with not invalidating them.
- How does LCP correlate with other performance metrics in the lab? Correlation study of over 4 million sites on HttpArchive Mobile and Desktop. Largest contentful paint correlates well with speed index and not very well with first contentful paint.
- Does framework usage affect correlation? Yes, the correlation is lower when using frameworks, as detected by Wappalyzer on HttpArchive sites. Scatterplot charts for top frameworks on desktop
- What about outliers where Largest Contentful Paint is very different from Speed Index? There are some examples here.
- Should elements be ignored as candidates for LCP if they are removed from the DOM? Examples of pages which differ depending on whether removed elements are ignored