Increase of TTFB in 2015 relative to previous years

Analysis
1

I was expecting to see a decrease in the median TTFB (time to first byte) for HTML pages delivered over time due to all these latest advances but to my dismay I see an alarming increase of late. Any theories as to why?

Here’s the query I used to generate the above graph:

SELECT  
  date,
  NTH(50, quantiles(TTFB)) TTFB
FROM (
  SELECT  STRFTIME_UTC_USEC(INTEGER(createDate*1000000), "%Y-%m") date,TTFB FROM 
	httparchive:runs.2014_01_01_pages, httparchive:runs.2014_02_01_pages,
	httparchive:runs.2014_03_01_pages, httparchive:runs.2014_04_01_pages,
	httparchive:runs.2014_05_01_pages, httparchive:runs.2014_06_01_pages,
	httparchive:runs.2014_07_01_pages, httparchive:runs.2014_08_01_pages,
	httparchive:runs.2014_09_01_pages, httparchive:runs.2014_10_01_pages,
	httparchive:runs.2014_11_01_pages, httparchive:runs.2014_12_01_pages,
	httparchive:runs.2015_01_01_pages, httparchive:runs.2015_04_01_pages,
	httparchive:runs.2015_02_01_pages, httparchive:runs.2015_03_01_pages,
	httparchive:runs.2015_04_01_pages, httparchive:runs.latest_pages,
) GROUP BY date, order by date;
2

Have you tried restricting the query by (Alexa) rank? As we continue to expand number of sites crawled and reach further down into the tail… I wouldn’t be surprised to find metrics “regressing”.

1 Like
3

You are right in that the delta shows up in Alexa rank greater than only 10,000

Less than 10K seems sane