Outlier inspiration

You have probably heard it by now. Facebook has acquired Instagram for $1bln. The best (humorous) reaction I have seen so far was chx's appeal that he will have to stop using Instagram (you can be forgiven if you are not familiar with Instagram, but not if you don't know who chx is). The worst reaction - and sadly, not because it was bad humor but because it was dead serious - was this tweet:

12 employees. 31 million users. $1 billion dollars. That's the kind of leverage focus gets you.

This is utter nonsense, and dangerous nonsense, at that. The implied lesson is that any small company could sell for a billion - as long as it focused. This is about as accurate as arguing that buying a one dollar lottery ticket can win you $100,000,000. The statement is true; only the real chance of it being true is so infinitesimally small that for all practical purposes the statement is false. The same applies to the 12 employees / $1bln dollar equation. False in an overwhelming majority of cases, true in one extremely rare uncharacteristic case. Statistically, a sample of one doesn't mean a whole lot - means nothing at all, really.

You could say this is a lousy comparison - after all, lotteries are random, and companies aren't - it's all about focus for Pete's sake, focus is what makes this one company rise above the rest and defy trends and theories. You could say, yes, this is an outlier; yes, this is unique case - so let us not turn up noses at it and write it off as a fluke, but study the success of it in detail and learn from it and try to replicate it.

But we are not offered to study it in detail here - we are offered one quick summary, a one-word solution - "focus", which for a start-up company of several employees translates into "work 80 hours a week, eat ramen noodles, sleep under your desk, and forget about exercise and life". This is a very lousy proposition, especially if your chances of success are small. At least in a lottery you can buy a ticket for a buck - you don't have to ruin your eye sight, your back, and your prospects of a family life to give it a try.

What is really scary to me is how often we are given tech outliers as a source of inspiration. As we get caught (in the academe and outside of it) in the technological vortex of me-too projects, we never look at the mean anymore - only at the top .0001%. For example we say "Apple achieved 500% growth in one year in such and such segment let's figure out how they did it" instead of saying "an average tech firm grew 3% in that segment, let's learn from them, and let's not forget to look at firms that lost share this year and see what they did wrong". We say "MIT has a great iPhone app, let's ape that" instead of saying "Most university mobile apps are a pain to use and an embarrassment; let's try to avoid adding ours to that list".

What you learn from outliers upon inspecting them is the combination of the unique conditions that have created them (and the realization that you don't have any of those conditions in your case). If you are going to look at numbers, your safest point estimate still comes from the measures of central tendency (mean, median and mode) and not from the new "science" of outliers.