What is the root cause of outliers by page weight?

A query to isolate top ~3% of desktop pages by page weight:

SELECT 
  rank, url,
  INTEGER(bytesTotal/1024) as TOTAL, 
  INTEGER(bytesHtml/1024)  as HTML,
  INTEGER(bytesJS/1024)    as JS,
  INTEGER(bytesCSS/1024)   as CSS, 
  INTEGER(bytesImg/1024)   as IMG,
  INTEGER(bytesFont/1024)  as FONT,
  INTEGER(bytesFlash/1024) as FLASH, 
  INTEGER(bytesJson/1024)  as JSON,
  INTEGER(bytesVideo/1024) as VIDEO, 
  INTEGER(bytesAudio/1024) as AUDIO,
  INTEGER(bytesOther/1024) as OTHER
FROM [runs.2015_12_01_pages]
WHERE bytesTotal > (
  SELECT NTH(97, QUANTILES(bytesTotal, 101)) as cutoff FROM [runs.2015_12_01_pages]
)
ORDER BY bytesTotal DESC

On quick inspection, most of the outliers are due to lots (or heavy) images, or large autoplay video files:

Running a similar query against mobile data yeilds:

Almost all of the mobile outliers are explained by lots (or heavy) images.

2 Likes