Skip to content

Death to Data Hogs! Ooops, that’s us!

The New York Times has an article on the concentration of use of mobile airwaves.  It seems that 1% of the users consume half the bandwidth:

The world’s congested mobile airwaves are being divided in a lopsided manner, with 1 percent of consumers generating half of all traffic. The top 10 percent of users, meanwhile, are consuming 90 percent of wireless bandwidth.

Arieso, a company in Newbury, England, that advises mobile operators in Europe, the United States and Africa, documented the statistical gap when it tracked 1.1 million customers of a European mobile operator during a 24-hour period in November.

Wow!  That is amazing!  Those greedy bandwidth hogs!  If it wasn’t for that 1%, usage would go down 50%. Get rid of 10% and usage goes down 90%!  Time to Occupy the Download Link!  Right?

Wrong.  As my former student Ben Peterson pointed out, measuring usage over a day doesn’t give much of an idea of the daily variability of usage.  For instance, three days ago I had some time to kill, so I downloaded a video and a few games.  I then updated all of my apps, and cleaned up my photo collection by sending some pictures to my web photo storage.   I was a complete bandwidth hog!  Since then, I haven’t used my phone much at all.  So, three days ago, I was part of the 1%;  since then I haven’t been.  Which day did you measure me?

So yes, on a given day, 1% of the people use half the bandwidth.  But, unlike financial wealth, we get to change around (every day!) who that 1% are.

If you want to use analytics, you really have to think about what your statistics mean and how you can generalize it.    The New York Times (again!), like much of journalism, rarely seems to take the next step and actually think about the numbers.

{ 1 } Comments

  1. Marcus Philip | January 8, 2012 at 4:49 am | Permalink

    It would be interesting to know if the average individual variation is larger than the variation in the sample. I don’t think so. I think there is a group of super users.