You might read a Data Scientist job description and get overwhelmed by the skills listed to do the job well. What they don’t tell you is your soft skills are just as crucial as being able to perform K-fold cross-validation. These are things you can learn and highlight in an interview that won’t completely make up on lacking experience with a specific technology, but still, help you be seen as a “great” Data Scientist in your next role.
Being able to explain to not only stakeholders but coworkers and customers alike is a huge asset. One of the biggest things you’ll have to do as a Data Scientist is to explain things to people. I have to explain Machine learning models, basic probability, why forecasts predict the way they do, and why correlations don’t equal causation. This also includes coordinating appropriate meetings with stakeholders do better understand the use case of models you create. Translating our industry jargon for the people around us and adjusting how we speak to technical and non-technical folks is a requirement in the modern Data Scientist's toolbox.
Time Management ⏳
Estimating project timelines is a huge part of my role. I can decide how much of our team needs to be involved and how long we think it’ll take to create and QA the deliverables. Whether this is an Excel document, Tableau dashboard, or Machine learning model, we scope out if it’s a small-ish, individual project or requires a team to work in two-week sprints.
Professional Writing 📃
You’d be surprised at the number of people who don’t know how to write well in emails. Being able to write a professional sounding email is a skill that can be learned just like any other. Truth is, we shouldn’t always write the way we speak, and we should be able to cater to the same insights in different levels of depth depending on who we’re talking to.
Technical Writing ✏
This mostly pertains to writing good documentation of your models and assumptions. We have to keep good, clear documentation of our work so it’s not only easy to reproduce, but when we go back to tinker with it, we have an understanding of what we were trying to do. It IS possible to write about your model or package in a way that’s not completely dry and hard to follow. Check out these tips on better documentation writing so it’s clear for users.
Tactful Critique 🔥
One of the hardest things about transitioning into data science for me was learning to be fearless when giving criticism and delivering it with tact. I had the tact down, but I was always worried about critiquing a company’s way of doing things. While that was easy for me as a people pleaser I had to learn that’s not the best thing for the business. I had to understand that if I didn’t push us to use better products and new techniques that I wasn’t creating the kind of environment that grows from its experiences.
File Management 🗃
Being able to manage your versions well locally is a big deal. Don’t be like me and prioritize speed for file organization. You’ll thank yourself later. Save often and come up with a system that works door you. Some people add a V1, 001, or “needs_cleaning” to the end of the file names to help sort through what changes you’ve made or which version you should send a colleague. Structuring your projects properly will help you stay organized and integrate well with any Data Science team.
Active Listening 👂🏽
You need to be able to decipher what your customers say they want and what they want. For instance “I need to see all customers that purchased though brand ambassadors last year” might really mean “I need to determine if the brand ambassador program is worth the cost to the company”. Many Data Scientists are in positions where they deliver data to other parts of their organization. This often means boiling down their work in layman’s terms, educating customers on their models, and framing how predictive models aren’t ground truth. To translate their needs into tasks you need to be able to ask the right questions that help customers reveal their motivations.
One tactic my colleagues and I employ in interviews is asking candidates what they would change about their past projects. We find insight in how people think if they’re able to critique their past projects and demonstrate they’ve learned new techniques. If you think there’s nothing you could have done to a make a project you've worked on better, it’s a big red flag. One of the best ways to do this is to detach your self-worth your work. You're not bad at your job because your model doesn't predict well. Take the extra step to understand why and even if you made something that failed or only predicted accurately for 12% of cases, you can explain what you'd do to improve performance or even better, call out how you may not have had the best grasp on the data and could have posed your hypothesis better.
Data Analysis 🔍
While this seems to be a hard skill, being able to stare at a table of numbers and decipher what's going on can be learned. I was practicing data science for years without stopping to really analyze my data. This happens by forming questions and building new biological neural pathways that act as a mental decision tree of questions to interrogate your data with. It wasn’t until I completed the Head First Data Analysis book that I fully understand how to pull insights from raw data. Without the pressure of a deadline or deliverable, I was able to study a small table of data and have an intuitive idea of what was going on.
Business Awareness 📈
It’s pertinent to know about the key performance indicators (KPIs) for your industry, especially in large organizations. Data science is useful not just in tech, but also in healthcare, marketing, and real estate to many a few verticals. Each industry has its unique challenges and the metrics for success are drastically different. Having a good knowledge of this wins you a lot of “nice-to-have” points while interviewing for a new role.
BONUS: A technical skill everyone in Data Science should have is an advanced knowledge of Excel. As much as we in the industry like to joke that Excel will never go away, it’s true. Maybe if you’re working at an AI-first or research company you won’t touch much Excel, but customers often still ask for deliverables in Excel workbooks. Advanced Excel usually means being able to perform V LOOKUPS, Pivot Tables, and Visual Basic.
Thankfully, the soft skills are a little easier to learn that becoming familiar with a new language. If you’re transitioning careers, you can practice a few of these in your current role and prepare to highlight your soft skills when you interview for jobs in Data Science.