Six must-have soft skills for every data scientist

Today, it’s hard to imagine a world without data science. Over the last few decades, it has become ingrained in society. Particularly during the COVID-19 pandemic, data is front and center in headlines every day. That being said, it’s important to remember that data science is still a new field, and one not without its challenges to overcome.

We surveyed more than 2,300 working data scientists, students, and academics in 100+ countries for our annual State of Data Science report, to understand how the data science discipline is maturing within commercial and academic environments. One concerning finding was that 40% of respondents said they “almost never” or “only sometimes” can effectively demonstrate the business impact of data science within their organization.

Without being able to show the impact, data science projects may not ever get off the ground — and those that do run the risk of being deprioritized or canceled. This hurdle often has less to do with technical skills and more to do with soft skills like communication and relationship building.

These six tips will help data scientists, and those who work alongside them, demonstrate the value of their work.

1. Prioritize the right projects

To be able to show business value, you must first identify projects that ladder up to broader business priorities. While a project may be interesting or valuable in another way, if it doesn’t help achieve identified business goals, it is going to be an uphill battle securing stakeholder buy-in and showing meaningful results.

2. Bring stakeholders in early

At the same time you are evaluating projects, you should also be identifying and pulling in relevant stakeholders. Successful implementations require the buy in, feedback, and resources from people in different areas of the business.

Spend as much time understanding your stakeholders and their needs as you do understanding your data — this is a step that many data scientists don’t spend enough time on. Having a thorough understanding of what decisions stakeholders will be making using your data will help you be a better partner.

Folks from IT and product are obvious stakeholders, but data scientists should go a step further by identifying executives with an interest in the project’s success and securing their sponsorship. Executives can help identify and engage other stakeholders, ensure a project is aligned with larger business goals, and remove obstacles.

3. Set clear KPIs and provide regular updates

After aligning on a project’s purpose and scope, the next step is to define how you will measure its success — or failure. There are KPIs that speak to a data science project’s progress and inherent success, such as different milestones or metrics like precision, target distribution, etc.

However, it is also critical to tie your project to KPIs that reflect business goals and metrics. This may require translating data science KPIs into language that internal, non-scientific audiences will better understand like cost, revenue, reduced churn, or time saved.

Once you have defined metrics, reporting on progress doesn’t end there. Communicate progress on your goals regularly; even better, create a dashboard or an application that stakeholders can view and manipulate this information themselves.

4. Achieve and communicate quick wins

As a project gets up and running, getting a quick win can be a crucial way to maintain momentum. A quick win means a goal that connects to business priorities that you can define, reach, and achieve within 6-12 weeks. Quick wins keep stakeholders invested and show your organization that you are delivering immediate value.

To do this, access to clean, readily-available data is a necessity; keep that in mind when identifying your possible quick wins.

5. Know your audience

As you may have noticed, clear, frequent communication is critical to showing the impact of your work. In our survey, nearly a quarter of respondents said that the data science/machine learning area of their organization lacked communication skills. Data science teams can often be isolated from the rest of the business and most comfortable using technical language that others may not understand. But data scientists have a responsibility to tailor their communications to their audience: use metrics that your stakeholders care about and language they understand. Use stories, analogies, and effective data visualizations to represent your data.

6. Visualize it

Data visualizations can help you more effectively communicate and show impact, but they can also distract, confuse, or mislead if used incorrectly. Similarly with communication skills, 24% of survey respondents said that their data science team lacked data visualization skills. Moreover, only 49% of the students who took the survey said that they were being taught data visualization in school.

Unlike R and Javascript, there are an overwhelming number of visualization libraries to choose from in Python. Pyviz.org is a great resource to learn more about different tools and how to use them. At Anaconda, we support the open-source visualization project Holoviz, which focuses solely on the tools that support web browsers. When you’re trying to communicate with business stakeholders and leadership, you need to work with a tool that allows you to share your visualization easily.

When selecting and using these tools, it is critical to ask what you are trying to achieve through the visualization. Be intentional about how you are communicating the data. Don’t try to force-fit your data to an inappropriate visualization.

With a clear plan, intentional relationship building, and good communication skills, those in the field of data science can better represent themselves, their work, and their business impact. And with increased understanding of data science comes increased investment and trust from the broader organization, crucial elements as the data science discipline gains maturity and prominence.