How many as$#oles should you have on your team, by way of normal distribution

Have you ever thought about the distribution of personalities in your technology team? What makes for a great technical team-member? I would like to focus on just two traits: 1. Technical excellence, 2. Not being a dick.

The first one is obvious, you want smart, capable, knowledgeable people to deliver good technology.

The second is important for a team to work together with good communications and collaboration. 'Being a dick' covers a broad spectrum consisting of drama-queens requiring endless attention, lone wolfs refusing to share with others, non-conformists rejecting team standards/procedures, brown-nosers selling the team out for their personal gain, and so on.

It would be a reasonable to assume that each of these traits have a normal distribution among the population, but what about the combination of the two?

Small digression: Normal distribution

In 1809 Carl Friedrich Gauss, a German mathematician with (apparently) a fetish for ugly hats, published a paper introducing the concept of Normal Distribution. The paper was an immediate hit, and the term “Gaussian distribution” was coined. This gave Gauss immortal fame, and totally pissed off his mother-in-law.

When measuring naturally occurring attributes (e.g. height of children, color of flowers, etc.) in a large population, most results will be close to the average, and the further out you go, the less results you get. If you plot out the value of the observation on the X axis and the number of observations on the Y axis you get the famous bell-shaped curve.

The reason for this behavior is that these attributes are caused by many uncorrelated random variables.

Take for example the ability to write good code. Let’s assume for a second that it was only impacted by one variable – mathematical ability – and that this variable is totally random (i.e. evenly distributed). We would expect that the ability to produce high-quality code will also be random and evenly distributed. We will get the same number of people producing eye-watering poetic code, great code, acceptable code, horrible code, and code that makes your eyes bleed.

Now let’s extend the model and assume that the ability to write great code is also impacted by another evenly distributed variable, say long-term memory. Most importantly, these two variables are uncorrelated (each is evenly distributed independently from the other).

Now the picture changes. There will be few people with high math skills and great memory, few people with crappy math and bad memory, and lots of combinations in the middle. This is akin to rolling two dice over and over and counting the results. You will get very few 2’s, very few 12’s and a lot of 7’s.

And if instead of only two variables there are tens, hundreds of even thousands of variables all mostly random and uncorrelated, you will get a large population in the middle and very few at the ends. Hence the bell shaped curve.

It is not impossible, just highly improbable.

Hitchhiker's guide to the galaxy, Douglas Adams

But…but…but…

We started with two traits, tech proficiency and not-being-a-dick-ness, and if these two are uncorrelated, you would expect each to be normally distributed, but the combination of the two to be evenly distributed. If you read this and think: “thanks for the statistics lesson, but that is not what I’ve experienced”, I tend to agree with you.

In my unscientific, empirical experience, there is a fat tail (i.e. large distribution) of people that are great technologies and also not very social. Perhaps its causal; maybe being a great technologist leads to arrogance. Or maybe both of these traits are caused by a third variable which makes them corelated. Who knows.

I knew it, I’m surrounded by assholes.

Dark Helmet, Spaceballs.

So what’s your point?

I once worked at a firm that optimized for technical ability to the extreme and encouraged an anti-social culture. Each individual was technically brilliant, but they could not work well as a team. There was no documentation, no unit testing, no coding standards. In short, no good development standards and processes. Maintaining systems there was a total nightmare.

And beyond the technical issues, it was just an incredibly depressing place to work at. Let’s face it, nobody wants to work with assholes.

So while we all want to work with great engineers, being a great team member is just as important. I would always prefer a good-enough (but not bad) engineer that communicates well and cares about the team over a rockstar engineer with no social skills.