SELECT js_bucket, COUNT(*) pages FROM (
SELECT
ROUND((bytesJS/1024)/50)*50 js_bucket
FROM [httparchive:runs.latest_pages]
)
GROUP BY js_bucket
ORDER BY js_bucket;
Above queries “buckets” pages in 50KB increments, which yields the following for Jul 2013:
I’ve clipped the tail in above graph - some sites are downloading 10MB of JS! But once you get over 2MB, it’s <10 sites per bucket.
SELECT
NTH(50, quantiles(bytesJS)) JS_med,
NTH(75, quantiles(bytesJS)) JS_seventy_fifth,
NTH(90, quantiles(bytesJS)) JS_ninetieth,
NTH(95, quantiles(bytesJS)) JS_ninety_fifth,
FROM [httparchive:runs.latest_pages]
So, 171KB is the median, and 640KB for 95th percentile (also for Jul 2013 data).