Several years ago I was speaking at an “Analytics Summit” that brought together all—or many in this case—of the analytical practitioners at a big American bank. I take pride in often speaking at the first such gathering a company has, but at this bank it was their second one. My host, a perfectly affable fellow with a Ph.D. in a technical field, told me this before the event started: Last time we invited both the heavy quants and the light quants [their term for analytical translators]. But this time we didn’t have room for everybody, so you’ll be glad to hear you are only speaking to the heavy quants. I was not glad to hear that. In fact I said in my presentation that I thought it was a big mistake to exclude the “light quants,” which is a somewhat insulting term to begin with. I said then, and feel even more strongly now, that translators are just as important as the most talented Ph.D. data scientist who can optimize an XGBoost algorithm. The analytical translator—a far better name than “light quant,” which defines people only in terms of what they are not—has gotten some deserved visibility over the last couple of years, in part because of a good McKinsey article. I wrote about the same type of folks a few years back, using perhaps even a worse name than light quants: “purple people.” Fortunately, I didn’t invent that term—I think Wayne Eckerson did in this 2010 blog post. What they do, as you probably know, is to translate business needs into analytical solutions, and translate analytical outcomes to business people. But it’s not really what you call these folks; it’s that you hire and respect them. As at the bank I mentioned, even if an organization knows that they need translators, they often don’t get the respect they deserve. Organizations seem to value people with deep technical skills more than those with business acumen, the ability to communicate complex ideas in simple terms, and a strong interest in solving business problems.
Not Just in Analytics Groups
I don’t know why organizations shower glory on heavy quants, but I’ve seen it in other circumstances, including business schools. There I call it “physics envy.” Business professors are envious of their more rigorous and scientific colleagues in the natural sciences, with physicists at the top of the prestige scale. So they try to emulate them. One might think that the people business schools would value most would be those whose ideas are absorbed and used by businesses, but that’s not generally the case. Instead, those professors who can solve complex equations are typically valued more highly than those who can solve (and communicate about) complex business problems. Those who teach algorithms are valued more highly than those that teach communications or consulting or coaching skills. Despite the lack of prestige, I’ve made my choice to be an academic translator and I’m happy with it.
What Happens When You Focus on Heavy Quants
What are the implications of over-valuing highly quantitative individuals and under-valuing highly communicative ones? One is that business leaders are likely to find the analytics or AI group difficult to communicate with. Some have suggested to me that highly quantitative individuals are less likely to be interested in solving routine business problems with analytics or AI. Others have said that hard-core quants have less interest in learning about the business and its customers. None of these issues are inevitable, but they seem to be pretty common. Another issue with an overemphasis on the highly quantitative is that they are the most likely group to be replaced—or at least heavily augmented—by automated machine learning. AutoML, as I have written before, can do many of the tasks that were previously done only by talented data scientists, including selecting the best algorithm, engineering features, and explaining the model. What it can’t do is just what translators do: engage with stakeholders, understand what data is available, explain the model in nontechnical terms, and address change management issues that hinder implementation. My guess is that the need for data scientists will shrink over time because of these tools, and the need for translators will only grow. I have also found that translators tend to be the best candidates to manage data science and analytics groups. They’re more likely to be interested in and good at managing people, and at reaching out to business leaders to evangelize for quantitative solutions. They may also be more interested in justifying the existence of quantitative groups to senior management. So if you are primarily focused on hiring and promoting heavy quants, you may end up with no good candidates for managing analytics organizations.
The Impossibility of “Having It All”
Of course, it would be ideal to hire people who have it all. If you can find an individual who is great at math and statistics, a fantastic programmer, an excellent communicator, replete with business acumen, and who deals with others in an empathetic and supportive fashion, by all means hire that person. Sadly, however, those people are unicorns. Those who received Ph.D.s in rigorous quantitative disciplines are unlikely to have done so because they love dealing with people and solving typical business problems. Instead, they generally seek intellectual stimulation, unique and hard problems, and a lot of autonomy. They don’t revel in explaining regression or decision tree models to people who don’t understand them. If you agree that hiring and valuing translators is important, there are many ways to show it. You can have different career ladders for data scientists and translators, but ensure that translators have the same potential for career growth and compensation. Highlight successful translation achievements as much as a new algorithm. Offer translators the opportunity to learn new skills, just as you might publicize and support a Python course. If you hire externally, try to identify great translators you might hire just as you would identify a prominent data scientist you’d like to bring in.
We’ve made substantial progress in realizing that translators are an essential aspect of analytics and AI organizations. Now all we need to do is grant them the respect and importance that they deserve.