If you read enough of the job postings, it will become clear that the term "data scientist" still does not have a clear definition. Some companies go as far as to ask for every possible skill and experience -- the so-called "unicorn" job postings -- and it's common to include academic degrees in that mix. That raises the question: Should any (or all) of your company's data scientists have a PhD?
We've seen this before. During the 1990s tech boom, most job postings for software developers (then, "computer programmers") required a computer science (CS) degree. They even demanded a master's degree or PhD for some higher-level roles.
That didn't turn out as planned. A CS degree did not prove to be a strong indicator of a person's ability to write software. Plenty of people who lacked the degree proved themselves to be talented, accomplished software developers. At the same time, plenty of degree-holders lacked the skills required for the professional software workplace.
Part of the problem was a cognitive disconnect between the term "computer science" (which leans toward data structures, developing custom algorithms, and research on new ways of framing deeper computational problems) and the role of a software developer (designing, implementing, releasing, and maintaining business applications). Both are important, but they are not the same role. What most companies needed were the latter.
(NB: Some academic programs now offer CS courses that better prepare students for the professional workplace. Twenty-five years ago, that was quite rare.)
Granted, there were companies that genuinely needed that advanced computer science knowledge. They were building high-end software that required speed and efficiency well beyond what was available in off-the-shelf libraries. (Consider, for example, the early days of distributed computing.) But for the most part, companies that required a formal CS degree were creating an unnecessary and artificial barrier to entry.
Sound familiar? We're seen the same thing in data science hiring today.
It's rare that that a data science job genuinely merits a PhD or other advanced degree. Sometimes a company's first data scientist holds a PhD and, as companies often do, the job posting reflects that they want a clone of that person for the next hire. Some do it for prestige, as it gives them the ability to boast that the entire data science team holds PhDs. More often than not, though, it's because the hiring manager keeps reading articles that claim a PhD is required to be a data scientist. And that is very far from the truth.
So, then, does your data scientist need a PhD? Probably not ... which means it's better to frame the question as: when does your data scientist need a PhD?
The answer depends on where your company finds itself on its data journey:
... it is highly unlikely that the data scientist role requires a PhD.
At this stage, you want someone with a well-rounded skill set. They should have robust technology and industry experience so they can lay the groundwork for your company's future data efforts.
(This, by the way, is why I often recommend companies hire a very experienced person as their first data scientist.)
Is it possible that someone with a PhD can fit this role? Absolutely. So long as they also have the technical skills and business knowledge mentioned above. Plenty of people fit the bill, and most of them do not have a PhD, so you'd do well to cast a wide net.
... it's still unlikely that you'll need someone who holds a PhD.
In this phase, your company is using off-the-shelf tools to analyze data and to build models, and you need more muscle to support those efforts. You may need to hire more experienced, well-rounded people (like your first data scientist) or you may need someone who has a specialized knowledge of the techniques and algorithms you're using.
In that latter case, you may be able to engage someone on a short-term project to build out some functionality and transfer the knowledge to your existing data science team. In some cases, this person holds a PhD and what you're working on is closely related to their doctoral research. And that word leads us to the third case:
... a PhD is still not required for the role, but this is your strongest case for asking for one in the job posting. That's because you're moving into the world of research.
Having outgrown the performance and problem set addressed by the off-the-shelf tools, you have to build something special. You need someone who can focus on a specific problem for long stretches of time, and who can can translate the knowledge and equations from published research papers into working algorithms to address your company's problem space. Plenty of people have this skill, though a PhD is a strong signal of this capacity.
In order to hire a person to do research, though, you also need confirm that your company can support a proper research effort. You'll need to have a longer time horizon on the outcomes of the work (you can't knock on the researcher's door every morning to ask why the probem isn't solved yet) and you have to realize that not every research project will yield fruit. If you are under time or budget pressure, then it will be difficult to support proper research.
The term "data scientist" covers a broad range of roles. Keep that in mind as you write the job description for your next data science hire.
Most data science work involves analyzing data and operationalizing the results. In this case case your searches should focus on candidates who have (or who can quickly learn) the skills required for the job. Academic pedigree doesn't play much of a role, so by limiting your search to candidates who hold PhDs, you're closing the door on a lot of talent.
In some specific cases, your data science work is more research-oriented. Here is where a data scientist might benefit from an advanced degree. Make sure your company has an environment and budget to support a proper research department if you go this route.
Whatever the case, be sure to tune the job posting to the role at hand. That will increase your chances of getting the right hire, sooner, which shortens the time till they are able to help your company meet its goals.
What should my company do with its data?
Executives want to know how to employ ML/AI in their company. They need more than just quick tips.
Why don't we talk more about risk in AI?
The AI world is sitting on all kinds of risk ... but no one wants to talk about it.