It is only natural to think of coding and everything technical when you say data science but that’s not necessarily the case. Even though data science requires you to have some technical skills, most of the time what counts is how much you understand the problem and how well you can explain your solution.
In this article, I will talk about what I think are the most important soft skills that you should have to be a good data scientist. Of course, it doesn’t mean that you should have all of these skills from day one, you can always work on and improve your soft skills. After knowing what to look for, all you need is practice
It sounds vague to say you need good communication skills. It is probably the most common criterion in job listings. It can mean a lot of things or nothing. But here, I divide it into sub categories for clarity.
have to communicate clearly with your stakeholders on a project and make sure to cover everything that is in the scope of that project before you start diving into the problem. One might argue that it is true for any project but I think it is particularly important in data science. It is not uncommon to have a fully functioning and finished machine learning model only to realise at the final presentation that it is not what your stakeholders’ were looking for. There are a couple of key points you should always make sure to have clarity on; the problem, the data and the deliverables. In order to do so, when you’re at the exploration stage of a new project, make sure to:
Get a good grasp of the problem before moving on.
The main reason people fail to do this is the difference in vocabulary and overall way of thinking you and your stakeholders will have. They will describe the problem from their point of view, which will likely be a bird’s view look at the situation in a business-y way.
Tip: To overcome the communication barrier make sure to ask all and any of the questions you have.
Have a good understanding of the technical requirements.
It is your responsibility to bring what your stakeholders’ are telling you into technical terms and understand what needs to be done with/to the data you are given.
Tip: I always find it useful to go back to my stakeholders, after I think I understood the problem, present to them my understanding and get their approval that I have fully and correctly understood what’s in their minds.
Understand the data and its pitfalls, corner cases.
Depending on the client/company you are working for, you might have more or fewer problems with the data. It is not uncommon to be receiving gigabytes of data with column names such as “ATT_BACK_LL” and without any nice looking explanation sheets. You might need to track down the people who will explain the data to you. Never assume the meaning of any information in the data you’re not 100% certain about. Getting insider information on the data is crucial to be informed about small caveats that might make or break your solution. I learned in one of my projects -luckily early on- that the timestamp attribute I was relying heavily on was off by 2 hours between certain dates while on a casual data understanding meeting.
Have clarity on what your stakeholder wants to see at the end.
Sometimes you might need to come up with it yourself from what they tell you but always double-check to avoid working towards a goal that doesn’t exist.
When it comes to data science I’ve learned that there is no such thing as too much communication.
As great as your solution could be if you can’t get the brilliance of your model across you’re not going to be able to satisfy your clients/stakeholders. Just like the issue of understanding the problem from what your stakeholders are telling you, you also have the responsibility to communicate your solution back to them in terms that they will understand. You don’t have to be a great presenter to achieve this. What you need to do is to make sure to:
Be on the same level with the stakeholders.
Using the vocabulary that your stakeholders are using and having empathy helps you look at your solution from their point of view and explain it to them. I know this sounds abstract but most of the time, you can see from people’s faces when they’re confused about what you’re saying, so you’ll know to explain something differently. Tip: Make sure that they know you are open to all questions and you want them to understand everything. A good trick I learned was to ask “What questions do you have?” instead of “Do you have any questions?”.
Have visuals to get your message across.
This can be anything from small PowerPoint animations to graphs and charts.
I don’t mean negotiating for a higher price or more hours to work on a certain project but simply negotiating the expectations of your stakeholders. Data science and machine learning or AI whatever you wanna call it is really popular and well known currently but still people don’t really know what to expect one day hear these phrases. Unfortunately, many people tend to think that some sort of magic will happen and older solutions will be solved in exactly the way that they needed to be solved and more. As a data scientist, you should make sure that you communicate the capabilities of your solution to the client and make sure that they do not expect a perfect solution to their problem.
Tip: Do not assume that your client knows what data science is and what it’s not. You can give them examples of the projects that you did, how they worked or even give them a small presentation of how machine learning systems/data-driven solutions work. After the exploration stage, I tend to go back to my clients present them my understanding of the problem and also a preliminary description of the solution. this helps me to make sure that we’re all on the same page.
I learned that one of the key skills you need is the ability to listen. What your client is asking you to do might sound silly or impossible but when you listen carefully, you might realise that even though they are putting it in the most complex way possible, what they need is as simple as 3 lines of code. Try not to discard anything out of hand and give what you’re hearing some thought.
Be open to receiving help. Once you start working professionally, it is really easy to think that you need to have all the answers. You might feel pressure to act like you know what you’re doing at all times. When you feel that way, remember that no one feels 100% confident in their skills. And if they are, it’s not a great sign anyway. Being vulnerable in this context keeps you open to learning opportunities. Bounce your idea off your colleague, ask for your supervisor’s opinion when you get stuck. You’ll be surprised how many things you learn and how fast when you let people help you instead of struggling with it yourself for days.
These are the 5 key soft skills, I think, a data scientist should work towards improving. They cover some of the common areas I’ve observed to be neglected by new data scientists. Of course, it is only human to not be perfect, especially when you’re new but keeping your clear-communication, presentation, negotiation, open-mindedness and vulnerability qualities in mind on your day-to-day journey will help you get better at them with every passing day.
On the flip side, if you think you are really good at any of these areas, make sure to make it known when you’re applying to jobs. Give examples of times you dealt with a hard customer, how you solved it or tell them about the really complex project you manage to simplify for your team. Soft skills matter, so if you have them—particularly if you don’t have many hard skills—highlight them.
I wish you luck and patience at dealing with your stakeholders in your data science career. You’re probably going to need it.