Feb 10, 2014
I attended a round table with Actian recently, one of a series it has been running around the world. Nominally on the subject of business analysis, it seems that at one of the US events, a comment that data scientists would become the next generation of CEOs seemed to get a warm reception.
Inside my own head, I could hear the screams of thousands of voices all saying “NO!”
The theory behind the comment seemed to be that all decisions should be based on better data and analysis, and therefore, the data scientist would be the optimal person to be in the post.
A couple of things wrong with this. First, I doubt that Carl Benz came up with the first diesel car based on in-depth analysis of spreadsheets; that Larry Ellison used someone else’s database and some massive analytics engine before deciding to found Oracle, or that Mark Zuckerberg sat down with a massive heap of data before coming up with the idea for Facebook. No — each of these were entrepreneurs, dealing with gut instinct and a nose for the next big thing — something that business analytics and business intelligence are not that good for. Sure, many “entrepreneurs” are serial failures; many are actually followers; and many make very poor CEOs anyway, but let’s look at the second issue.
We’ve had rock star employees before — particularly in IT. We’ve had the web app rock star; the web site rock star; the social networking rock star. Like all rock stars, eventually their star will wane: sometimes pathetically so. Having in-depth knowledge of a single environment is dangerous — particularly for a CEO. And data scientists — of the wrong sort — will be in-depth nerds of the first order.
Don’t get me wrong. I am a fully-trained and practiced scientist: I spent my first 10 years working in research for car catalysts, anti-cancer drugs, and fuel cells, amongst other things. And I therefore feel that I have more of a reason to fear true scientists in the business. The act of science is far more attractive to us than the outcome — particularly the business outcome. Does your business want to be run by someone whose discussions with prospects and the markets is likely to run along the lines of “When I looked at the Bayesian probability of a favorable outcome within the Jackknife variation against the variation over time with some pivotal z-score values and categorical variables, it became apparent that I should have no more than one lump of sugar in my coffee”?
A good scientist will be very good at their job. That job is to posit a position and test against that position to see whether it stands up — or fails. Very few true scientists will give you a 100% answer to anything. How certain is it that the sun will rise tomorrow? Not 100%. How many legs does a person have? Less than two, on average. Should we take this decision in response to what’s happening in a complex and dynamic market? Wait for a few months while I hit my data.
The real key is to try and encode the capabilities of the data scientist into ways that mortals can make easy use of them. Let the impetuous but clever entrepreneur test their idea extremely rapidly against as much available data as possible. Let groups of business people work together as a community using different types of data analysis to come to a more consensual agreement based on multiple readings of the same data — but all based on valid statistical approaches. Let’s embrace big data in its truest form — volume, velocity and variety — to maximize the value and veracity of the findings.
But don’t let a data scientist take over the company — unless they also can be seen to be business- and finance-savvy. If they are, they probably wouldn’t want to work for you, anyway.