You Ask, I Answer: How to Hire a Data Scientist?

In today's episode, Manina asks, not every marketer can or wants to be the data scientist for their organization. What should we look for when hiring an FTE or partner? So this is a critical question because there is a significant shortage of qualified data scientists, right? Those who exist and are qualified are, in the words of my friend Tom Webster, reassuringly expensive. Uh the last public figure I saw, there were something like 10,000 qualified data scientists with four years of experience or more in uh in America, and something like 15,000 marine biologists.

So more people know about whales than data science. So what should we be looking for? Well, remember, and we'll bring this up here. Data scientists have these four skill sets, right? Business skills, scientific skills, mathematical skills, and technological skills.

Here's the thing. Because of the talent shortage, the likelihood that you can find someone who is proficient in all of these is pretty rare. They are, you know, unicorns in a lot of ways. And so there are well, it's it's more likely that you're going to have to either find someone who has a strength in a couple areas and and is okay in a couple areas, uh, or uh more realistically, either hire an agency partner or hire a couple of people, or maybe even a team of people, depending on your budget, who can bring the different skill sets to bear. So this is the six sets of skills that you would be hiring for.

Now, presumably s you would have someone who's a business expert in your line of business uh and a domain expert. You uh someone who specializes in what it is that your company does. So those two should already be baked in. If you don't have those in your business, that's a bigger problem than a data science problem. That's a uh a fundamental business problem.

Data engineering is a set of skills where of someone who can work with where data is stored, how to store it, how to make it accessible, how to make it uh reliable. Those are uh critical skills, and that's what an area where many data scientists actually do fall pretty short because you need a fairly strong technology background for that. That's working with SQL databases and NoSQL databases, uh, graphing databases, uh, cloud, all the different cloud services that are out there. And data engineering is something where that person doesn't have to be on a data science team. They can be part of the IT team and then just be a resource that's available to be shared.

Because those skills are highly transferable to and from IT and uh are necessary in in most modern organizations. If you if you're storing data, you either have or should have a data engineer or someone with data engineering skills. That leaves the three uh primary areas for data scientists stats and math expertise, uh, statistics probability, linear algebra, some calculus, uh being able to understand the different theories and the different uh techniques and what they do and what they mean is critical. What they mean part is critical because it's easy to learn, you know, uh a particular mathematical technique. The harder part is pairing it with that either domain or business expertise to say this is why you would use this technique in this situation.

You have the coding, of course, in languages like R, Python, SPSS, whatever the statistical language uh of your choices to do data science work, and then that scientific mindset, that ability to uh adhere to the scientific method to set up well controlled tests and experiments, uh, the ability to understand that uh certain types of validity to understand like yet this is a valid experiment or no this is not a valid result uh being able to interpret data well from a reproducibility perspective those are the six core skill areas that you would be hiring for again it's difficult to find this all inside of one person that does it equally well everyone uh out there is going to have different levels of skill for example um I am weaker on the stats in math than I am on the coding I'm a better coder than I am a stats person. I can do most stats uh and I'm familiar with a good number of the techniques but I know that academically that's my weakest spot data engineering also not no problem. I know other folks who are phenomenal statisticians coding's not their thing uh and they have many many tools you know like at the SPSS modeler that allow you to uh circumvent that to some degree but if you're trying to work with the latest and greatest for example in you know neural networks uh you you do need coding ability one thing to be careful of when we're talking about hiring um you will need help interviewing if you don't have some background in these areas too because there are a lot of folks out there who did uh these crash course data science programs, right? Learn data science in six weeks. It's like saying learn surgery in six weeks.

I mean, yeah, you could probably become minimally competent at something, but it's not somebody I want working on me. Uh and there's a lot of those folks who are who are proclaiming themselves labeling themselves data scientists after going through one of these courses. You will need help interviewing to know what to ask people in each of these domain areas, to be able to ask them challenging questions, um behavioral questions, situational questions. Hey, you have this type of survey data that came back. How would you handle it?

Knowing that you can't go back and and redo a survey, or uh we have this data from social media, how would you interpret it and be able to assess the validity of their answers? That's gonna be the hardest part of hiring. I would recommend in a case like that, find a reputable data science agency and ask to partner with them just on the help us hire somebody who's qualified front. Um again, be real careful with those crash course folks. Generally speaking, someone who comes out of a data science crash course is going to have one and maybe one and a half of these these six skill sets, and more importantly, limited or no practical experience.

It's super easy to go through uh a training course, right, and follow the instructions, take the data sets that are provided in the materials and and work with them. And that's okay, right? That you you need to start somewhere. But if you're hiring for your company, you want someone with experience because expertise in data science, like all forms of expertise, is less about knowing how to do something when everything is great and the data is perfect, which it never is in reality, and much more about knowing what's going to go wrong. When you look at a data set and go, Oh, I know exactly what's gonna screw up on this data set, right?

And and exactly what you need to do to mitigate that and still be able to get your work done. It will never be perfect. But knowing, like, okay, what level of risk is acceptable here? Example. You have a data set with you know 25% missing data.

What technique do you use to manage that? And part of that question that experience teaches you is what's the level of risk? If you're doing if you have this data and you're doing it for like a marketing white paper, the risk is pretty much zero. Um so you can do something like predictive imputation, right? There's you're not going to kill anybody.

On the other hand, if it's a if it's a trial pharmaceutical that's going to be you know put into human trials, uh-uh. You don't you don't do the just let's fill in the missing data with a with a best guess algorithm, because you might actually kill people. Uh so that level of risk is substantially higher. So that's where those crash course folks, they yes, start them out as like a junior analyst, let them get their skills, but don't hire them as a data scientist and expect to get the same level of results that you would get from somebody who has the scars and the and and and the stories in order to deal with the things that are going to go wrong. So it's a really important question.

How do you hire and bring on these folks? If you're looking at an agency, again, assess these things, asking that agency, hey, how would you handle this? Be very careful if an agency proclaims they have data science expertise and you never actually get to talk to the data scientist. It's like you know, hiring a technology company, you don't ever talk to the engineers. Always talk to the engineer, always talk to the scientist.

They may not be the most personable people, um, but you'll get a sense very quickly of how good they are or are not in reality. It's a really good question. Uh if you have follow-up questions, leave them below. Subscribe to the YouTube channel and the newsletter. I'll talk to you soon.

Take care. Want help solving your company's data analytics and digital marketing problems? Visit trustinsights.ai today and let us know how we can help you.