8 Habits of Highly Effective Data Scientists

I’m fortunate to have met with some of the pioneers of data science and machine learning early on in my career. Their thoughts shaped my interest in the field and their habits formed my daily routine. The most frequent question I’m asked is some form of, ‘How do I build a machine learning or data science career?’ It starts with forming some important habits. Here’s what works for me and what I’ve seen build exceptional data scientists.

1. Focused Reading & Learning

There are only so many hours in the day and so much more garbage than gold to read. Develop a list of trusted sources. These are blogs, social feeds, and document aggregators that provide quality content. Don’t waste your time reading the fluff pieces you find on sites like Forbes or HuffPo. Don’t read every deep learning paper on Arxiv.

I follow the research being done at companies like Facebook, Google, Disney, and several others. They are frequent publishers and their work is worth reading. Universities like Stanford, MIT, Cornell, and others are also strong sources for quality research. Find bloggers who talk about topics of interest to you in a format that is easy for you to understand. Twenty different people all write/talk/video about the same topics so you have your pick of communication/presentation styles.

2. Optimize Your Workspace & Toolset

I’m faster than most data scientists because I have streamlined the tools I use. I can deliver projects sooner because I’m not wasting any time fighting with my development environment.

I have default server images on Amazon for my workspaces. Each one is customized with the IDE/environmental variables I’ve found best fits the programming language and database(s) I need to use. I have go-to data sources for most types of projects and prebuilt hooks into most internal data sources. Security settings are part of the image.

It took me a lot of trial and error to get here. I’ve used more configurations than I can count and experimented with several different data science and development applications. Build what works best for you and enables you to be most efficient. You will look like a professional in a field of unprepared amateurs.

3. Listen For Business Problems

If you don’t want to be building attribution models until Google automates you out of a job, you need to find new business problems to solve. I listen to Bloomberg Business and CNBC frequently. What I’m keeping an ear out for is why companies missed revenue targets. Those are the business problems they haven’t solved and are willing to pay for solutions to.

How can data science or machine learning tackle these challenges? I see a lot of predictive problems; something happened that the company didn’t anticipate. That’s often a supply chain disruption or a change in customer preference.

I also see data science or machine learning capabilities issues; a company doesn’t have the capabilities to analyze their data. These are problems shared by a lot of companies. Learn how to solve these problems and advertise your abilities. You’re a lot more valuable to a company if you can listen for their business problems and synthesize solutions.

4. Build A Professional Network

Reputation and influence are the new marketing. There are two benefits that make the work worth it. First, a professional network allows you to find people to learn from. Second, a professional network allows you to build a brand and grow your influence in the data science or machine learning community.

Social sites like Twitter and LinkedIn, as well as technical sites like Stack, are great places to build a professional network. Start out by following and listening. Once you start to see the types of content that the community likes to consume, move on to being an aggregator. Aggregating is as simple as sharing what you’re reading that your followers might appreciate seeing as well.

Begin to find your voice. As you see interesting topics that aren’t well covered by others or you have interesting experiences to share, start creating content. With Stack, start answering questions. As you gain expertise, think about contributing to open source projects or publishing your research.

Influence brings opportunities. Rather than cold calling for new jobs and promotions, you’ll be constantly promoting yourself. I’ve gotten speaking gigs, insider access to conventions, and several clients through my professional network. It’s well worth the effort.

5. Take The Time To Speak

Take every opportunity you can to speak. One of the biggest reasons I’ve been called an influencer is my speaking engagements. I take small, private audiences. I prepare a short talk on a topic of interest to the group and spend the rest of the time answering questions.

We all provide value in different ways. My style is to feed off my audience’s curiosity and direct the event towards important points around their areas of interest. Other speakers will spend most of their time on the presentation. Many audiences don’t have their own questions so it’s more valuable for the speaker to take them on a guided tour. This is better suited to a conference setting than a private event.

Develop your own style, message, and audience. Share your unique experiences, projects, or vision. You’ll be amazed by how many people are interested in what you can share and teach them. Speaking helps develop your unique perspective. In a long career with machine learning or data science, your perspective is far more valuable than any other contribution you’ll make in code or algorithm.

6. Say No A Lot

Data scientists and machine learning experts can do a lot of different things. That leads to us getting asked to do a lot of different things. In most cases, no is the right answer. Saying no a lot really comes down to understanding what you want to be working on and choosing your professional path. I’ve said no to jobs, book deals, and projects that didn’t fit into my personal path. I’m happier and more focused for it.

There’s a Venn Diagram out there with one circle labeled “right for you” and another circle labeled “right in front of you.” The overlap is very, very small. I’ve found that few opportunities that land in my lap are right for me. The opportunities I want, I must chase after myself. I’m the one asking for them, not the other way around. The most effective data scientists I know go get the projects, clients, and roles they want. I emulate that behavior myself and it’s worked out well for me.

7. Adopt A Minimalist Style

The masters use the fewest lines of code, the least data, the simplest algorithm, speak briefly, and so on. Minimalism is the mark of an expert data scientist.

8. Speak With Purpose & To Be Understood

I spend only about 20% of my time communicating with data science and machine learning experts. The majority of my communications are with non-technical audiences. They don’t care about the machine learning. They have an outcome they want.

The lion’s share of speaking with purpose and clarity is first listening. The process of asking questions to get to the truth of what a person or group needs is an art form. I’m still working on this one myself but I’ve seen the masters dissect a problem down to its root by asking the right questions.

The elements of a good dialectic are creating an environment where people are comfortable answering questions honestly and admitting what they don’t know, giving them a sense that spending the time to answer these questions will benefit them, and synthesizing their answers back to them in a way that shows comprehension.

Once I understand the question, I can answer it with a lot more certainty. I use language that anyone will understand while respecting my audience enough to expect they will grasp complex concepts if I can frame them the right way.

I see it as my responsibility to express ideas and concepts in a way that my audience can understand. When they’re lost or confused, that’s my fault, not theirs. This is the piece of my own advice I find hardest to take but that perspective on communication has helped me improve greatly.

I’m interested in your habits. What practices have made you a better data scientist? What have you observed in others that you work to emulate? Share your thoughts in the comments section.

Written by Vin Vashista.