How to Build a Data Science Team 

December 1, 2020

An experienced Vice President Of Data Science, diving deeply into the core foundations of building an amazing data science team.

The Data Science industry has grown significantly over the past 10, 15 years, and the demand for talent has gone up. When I started out, I was the only data scientist out there. Then I started leading small data science teams and then growing them out over the course of time. There’s always the need on top for more top talent. And retaining that talent, because that’s what drives everything.

As a Data Science team, we’re telling computers or teaching computers how to think. Without good people on your data science team, you’re never going to get results out. 

Welcome to The Tech Show. I’m Aziza. I’m here today with Focus GTS. Focus GTS is a niche IT staffing and recruiting firm, focusing on four areas of technology, Data Science & Artificial Intelligence, Digital Marketing Technology, BI & Analytics, and Cloud Technology.

We’ve partnered with you to bring in-demand tech talent to your organization, and fill those hard-to-fill roles. Today, we’re here with Andrew Eichenbaum. Andrew is a VP of Data Science at Egon Zehnder. He’s a leading data scientist, specializing in artificial intelligence and machine learning of data-centric projects, and leading teams for over 10 years.

Welcome Andrew, and thank you for joining us here at The Tech Show.

In this episode, we want to discuss building your data science team. Where do you start and what to look for. But before we dive right into it, I have to ask you.

Andrew, PC, or Mac?

Oh, PC. I am actually, I used to dive into Linux a lot in the early days and do a lot of my own building. I am very much an open ecosystem type of person. 

Essential Skills to Look for in Your Data Science Team

Let’s talk about the data science skillsets. The term “data scientist” expanded and became too vague, in recent years. Are you being an early member of the market, which skills are really necessary and which aren’t? Let’s talk about soft skills and technical skills.

It’s not an easy answer because there are a lot of different types of data scientists out there. There’s the person who is, for example, the Algorithms Person who delves in. But there are also other people, sort of the Data Wranglers, they will take 20-30 different data sources and put them all together in one view. And there is an assortment of other pieces. And I find that the best data scientists have a piece of all of these skill sets together. So it’s not easy to say these are the essential skills or these are the nonessential skills because it all goes into building your team.

Do you need somebody who will help pull out the data? Do you need somebody who’s algorithmically based? Or do you need somebody to understand and define metrics for the rest of the firm, which is not as simple as you might find in dashboarding? But all of that being said, there’s one thing that everybody has to have, and that has to be the ability to communicate.

Data Scientists are storytellers. We tell stories with data, and if you can’t communicate those stories out and have people see why there’s value in the work that you’re doing, it’s never going to be implemented. So what’s the use. 

How to Structure Your Data Science Team 

So when it comes to building your team, where do you start? How do you build an effective data science team that can produce? And do you start, from the top down on it from the bottom up?

It varies because very frequently when I step into a role, I’m very much zero to a 90 person. I’ll either get a small team of a handful up to maybe a dozen people or be the very first person to be asked, “can you build this out”?

It depends on the situation and where I go. 

If you have a team already in place, you have to evaluate where their strengths and weaknesses are, where they need help. And then build off of that. They could be top-heavy. They could be bottom heavy. Hopefully, it’s a nice distribution, but you never know. When I’m starting to build out a team on my own, I actually start in the middle.

It sounds sort of odd, but for the first two, maybe three hires, I’m looking for people with at least four or five years under their belts. So that I can hand them a project and say, go, and that just works. They can do the work on their own. They can use me as a resource for various pieces. I can continue to develop the story within the house as well as continue to hire and help them along.

But if we start at the bottom, I spend all my time guiding and mentoring people. And if I do have everybody at the top, there aren’t as many people actually getting the work out there. How the team then rolls out is really what comes across your door. 

Biggest Challenge When Building Your Data Science Team 

We speak to a lot of people and I’ve heard that many, many leaders actually have different strategies when it comes to that, and they all have their own thing to say. Some say you have to start from the bottom up. Others say you have to, especially in startups, have to hire somebody that’s up top and then have them build out the rest of the infrastructure of the team. So, you being the VP and being an industry for 10 years, what has been your biggest challenge when it comes to building a data science team?

Oh, I can’t say it still has to be sourcing. Simply as data science has grown over the past 10, 15 years. The demand has gone up. When I started out, I was the only data scientist and then I started leading small teams and then growing out over the course of time. There’s always the need at the top of that, and for more talent and retaining that talent because that’s what drives everything.

The data science team is responsible for telling computers or teaching computers how to think. Without good people on your team, you’re never going to get results out. 

I definitely agree with that. Sourcing is definitely, I think the hardest part as being in recruitment as well.

How to Retain And Keep Your Team Engaged

There’s a lot of people with the data science term on their resumes and their profiles. It’s definitely is getting saturated as far as the titles go, right? So how do you deal with those issues that you mentioned previously, like engagement and retention? How do you keep your data science team engaged? 

It’s a bit easier for me as I’m a practitioner and I still keep my hands somewhat dirty. I still do code reviews. I still work with my team at the whiteboard suggesting approaches, and it builds confidence in the team that when I represent them at the highest levels that they have somebody who knows what they’re talking about. So that helps me a lot. As for retention, that’s always hard. You need to be able to build a group in a dynamic where people are happy where they’re doing things. Data scientists can always sit in one place for a year, and two years later run off and get another $50,000 plus dollars.

And that’s going to happen the first couple of years. When you go from that $80,000 to $130,000 to $200,000 per year. Two or three years go by and this is going to be the way it is. But it’s really creating a team dynamic and having people work on interesting problems that really drives the data scientists. I find most of them are curious people.

They join a place because they’re interested in the problems presented and the team dynamic that they see where they believe in the vision. And so as you’re selling that vision, you have to make sure that that continues on through. If that team becomes stymied by somebody or, well what that dynamic changes, that’s when you see the real turnover happening. 

That’s interesting to hear because that’s something I’ve also heard from a lot of our recruiters is that at some point data scientists, especially the ones that have been kind of seasoned and been in the field for a while, tend to look for something that’s challenging and interesting now, right? So they’re looking for more than money, more than just compensation. And I guess from our conversation, the initial challenge besides overall scarcity.

The Importance of Selling Your Vision to Your Data Science Team

How do you address, kind of getting them interested, right? How do you sell that? Not every company is Amazon and Facebook and Netflix and so forth. So, I guess, how do you sell that vision?

When you’re at a startup or even a midsize company, there is a real target that you’re aiming for. An area that you’re specifically in. And when you’re talking about your Amazons and your Facebooks, they’re so large that they cover a wide range of pieces or a wide range of verticals.

So having people apply who are truly interested in the area that you’re looking for makes it so much easier. So I was one of the founders of Yummly. It’s a recipe search engine. And every data scientist, pretty much everybody we hired onto that company, was a foodie at some level. Whether they just enjoyed food to the point of cooking it and trying to improve upon their techniques in one way or the other. It was a real cohesive piece and that we were centered around food and those sorts of things.

If you can drive that within your group, that people believe in what they’re doing.

Culture is I guess, number one, it’s always all about the culture and people love food. What can I say?

The Combination of Hard and Soft Skills 

My next question would be what we’ve run into is that data scientists are extremely smart people. And it’s hard to kind of find both right. Somebody who is smart and good at building models and then somebody who can communicate that data into simple English to explain to the stakeholders and higher-ups. So how do you do that? 

It’s a tough skill. There are definitely people within the data science realm who are never going to be that person.

They like to stay in the data and like to stay technically savvy. And you have your data engineer tracks to sort of staff personnel. But for the people who want to go through the management and help, actually do a lot of the product management. Not just the personnel management, but the product management, it becomes that much easier for a project.

So it’s teaching them. From the point that they hit, they get out of that junior level. I like to have them present their work to the people above me. You know, as high as you can go, I’m more than happy to coach them, whether it be on a presentation or the actual talk that they’re giving. Making sure that they know at what level they’re at, and being there for them even during those first presentations. If they get flustered, they will have somebody that they know that they can lean on. 

But again, it’s just development. One of the things I look for, it’s one of the first questions I ask when I’m interviewing somebody. Is to explain a problem to me in a way that anybody who is a middle schooler or a high schooler would understand. And if they can’t do that, or they’re not interested in learning that, then there’s really no reason to move forward. 

The Importance of Communication Skills in Data Science 

How you look at data and can effectively communicate it to somebody who can understand it, that it is in middle school because it is a real thing. Even in marketing, you have to communicate your service or whatever you’re selling or whatever your company’s trying to do into simple English that anybody can understand.

I usually refer everything back to, call it “lift”, when it comes to data science work. One of the beauties of working with numbers is that we can always use those numbers to quantify what we do and why it’s better. You should be able to tell anybody, middle schooler or high schooler or anybody at the company you’re working at, and say, “I’m working on this project”.

They go, “Oh, that’s sort of neat”. You give them a two-minute description, and It’s like, well, “what’s going to happen?”. You say, “well, we’re going to try and deploy it or put out this algorithm metric, new dashboard, and we expect it to show something or create some sort of a lift. Whether it will be saved on the bottom line and improvements and the top line, better retention, or better acquisition”.

But you should have two or three metrics that you say, this piece of technology will drive, and that people can understand. You need to make sure people can grasp that. And it’s the same with anything else. With a front end system. You’re not just going to say, “build me a front end, a new user interface”. It is not like that, there’s a discussion that goes along.

There’s an understanding of what drives it, and It’s the same way with data science. It’s black magic, which I fully appreciate, but you need to have confidence. Your stakeholders and product managers, who might be working with you, need to have confidence in what you’re doing and once you have repeated successes, they’ll be like, “Okay, that’s great. Let’s move forward”. And there’s more trust built up and less discussion is needed. 

How To Keep Your Data Science Team Motivated

When you have a team and you have the stakeholders, obviously you have to produce results and so forth. How do you keep your team motivated and inspired as a leader to keep moving forward and making them feel like they really are part of the business and really moving the needle?

Part of it is that product understanding. Much like any good developer, you need to understand while you’re building something. And it’s the same with data science or a data engineer or an analyst. You need to understand the business value of what you’re driving and while you’re doing it. It was just like, “well, I see a spec I’ve written here, it’s been written here, I got to meet the spec”. There’s no engagement there. 

That’s how sort of boredom or apathy rolls in and you run into real problems then. Because you’re just looking at a paper, you’re not interacting with people, seeing what they really want out of it. This is why I like to engage anybody within my team, above the junior level with stakeholders, so that they get a real understanding.

They can start being part of the dialogue. They can see where the work is going and if problems arise, they can hear it directly from the stakeholders and they can work on it themselves.

Data Science Vocation

What is your favorite thing about data science and what would be your advice to somebody trying to climb up the ladder and someday maybe be in your position that you’re in right now? 

I’ve always had, at least for myself, I’ve always asked the question “why?”. Why does something work? I did my Ph.D. in physics because I wanted to know how the universe works. And I quickly realized as I was finishing out, there are lots of interesting questions out there besides the call it “STEM world”.

I was at Stanford doing my research and I got sucked into Silicon Valley. I’ve been able to answer all sorts of interesting questions there, you know? How do people approach food? What’s popular, what’s not? What changes seasonality? It’s always having those sorts of questions in your head that you’re trying to answer.

Advice to Climb Up The Ladder

Finding a position that will help you answer those questions will really drive you and keep you engaged on an individual level. As for the question of what skills? Whether there are hard or soft skills for anybody as a data scientist? 

For the beginning data scientist that has to be SQL. You need to know how to get your own data out and be able to look and understand along with a basic understanding of statistics. You need to know that if you’re flipping a coin so many times or throwing dice, what are these probabilities? Because this is the everyday life of understanding if a data scientist result or a generated result makes sense. 

As people get older, I look for production-level programming, which I guess is the best way. The stuff that we write just gets rolled out. If it’s inefficient, if it’s wrong, if it’s buggy, that’s going to cause lots of problems. You need to have a sort of production level coding abilities. Not everybody can reach that, but it’s one of the places I suggest. 

And I guess the next level is product management. It’s hard to find good data science product managers or any good data product managers. So data scientists very frequently take on the role themselves. So learning how to communicate, learning how to present, not just your final results, but how to have those interactions with your stakeholders to find out what they really want. And these are sort of the biggest things that I talk to within my team.

How To Contact VP Data Scientist Andrew Eichenbaum

Thank you for sharing that with us, and for our viewers, what is the best way to connect with you? Is it LinkedIn or how do you prefer I guess? 

LinkedIn is the best way and you can link up with me there, or, you know, my email address is just my name, andreweichnebaum@gmail.com. Feel free to drop me a note. And hopefully, it won’t get put into the spam box with the random other things that get put there that I need to move back.

Yeah, definitely. I’m also, we’ll be attaching a link to your LinkedIn profile right underneath this video. So, anybody who would like to connect with Andrew, you can go ahead and just click over to his profile and go ahead and connect with him. And thank you so much, Andrew for your time. And thank you for coming to this Tech Show.

We’re really thankful to have you. And it was a really nice conversation.

Thank you. My pleasure.

If you are interested in participating in the Tech Show of our IT staffing and recruiting agency, please feel free to contact us!