Found in 3 comments
by folbec
"Proofiness" :

It is also a common tread in books by Tufte :

I like "Visual Explanations: Images and Quantities, Evidence and Narrative" :

Original thread
by ChuckMcM
Sigh what a squishy article. What is the methodology? What is the sampling error? What does "OK" mean relative to "Good" ? This reminds me of the anecdotes in "Proofiness"[1] which talk about how the news industry figured out they could manufacture news by taking a poll.


Original thread
by ChuckMcM

TL;DR version the 'unlimited guys' use the same total bandwidth as the 'limited' guys.

This is extreme proofiness[1]. You notice that the 'average' is much higher in unmetered plans while the median is the same? If you know your maths you will recognize that data hogs hog a lot which results in a higher average. AT&Ts throttling on the other hand keeps the 'median' about the same.

The real issue for the carriers is that they have a fixed amount of bandwidth within their system which is difficult (read expensive) to increase.

Network bandwidth is one of those commodities that exists all the time, whether or not it is being used.

People don't predictably use the bandwidth so its hard to sell "all" of it for money.

So you end up with bandwidth that is sitting there unused and losing money (it has a constant cost of ownership which is an amortization of maintenance, depreciation, and licensing fees)

Their solution was to offer an 'unlimited' plan which, for an extra fee, lets folks go over the max tiered plan. This allows them to 'sell' bandwidth that might have otherwise gone unused. But if they drive the total use over their hard stop EVERYONE runs out of juice.

So their 'fix' is that if you're pushing the average up they clip your usage. That keeps the median in line with their highest rate plan while allowing them to sell the excess bandwidth they have for those cases when everyone asks at once for it.


Original thread

Looking for a good book? Subscribe to the weekly newsletter.