Data scientist roles are still not very common in HR departments. A successful establishment of such a function requires balancing the analytics maturity of the business and HR leaders with the data scientist’s skills. Therefore, it is essential and fascinating to explore how data science and HR needs are knitted.
I was fortunate to interview Elizabeth Esarove, a data scientist in a senior advanced analytics role in the HR department at AT&T. I contemplated many things yet to be learned and developed and found some clues for data scientists’ development between the lines.
A career path from HR roles to data science
Littal: Tell us about your background, current role, and professional journey from an HR professional to a data professional.
Liz: I was hired into an HR role due to my analytics experience in the banking industry. At that time, HR departments were beginning to understand the value of using data to measure and improve HR functions. My role was on the new data analysis team; I was the third person to join the group. That was about 20 years ago. Since then, I’ve had roles within HR that involved data analysis for Talent Acquisition, Talent Management, and Training, where I learned about each area of HR and improved my ability to provide valuable insights.
When self-service reports became important for HR, I moved into a project management role where I served as a liaison between Information Technology and HR to work on projects developing reports for HR teams. This role helped improve my ability to communicate effectively in business terms with HR leaders and technical terms with Information Technology. In addition, the experience helps in my role as a data scientist today as I explain the results of my analysis to executives.
In 2013, I read news articles about companies that used predictive models to reduce employee attrition. I could envision other potential applications for predictive models to address workforce and business issues. I was also ready to take on a new challenge in my career. So, I decided to get a master’s degree in Data Science to develop my skills in predictive analytics. One of my electives was a text analytics course which is an asset in my current role because it’s necessary to analyze comments received in our employee surveys. I started applying my new skills the year before I graduated and have been working on predictive modeling and text analytics projects for the past six years. I enjoy applying my data science skills to HR projects!
The role, skills, and tools of a data scientist in HR
Littal: How does your data scientist role differ from other roles in the team, such as data engineers or data analysts?
Liz: As a data scientist, I focus on statistical analysis, predictive models, or text analysis to provide insights for business decisions. Other roles on our analytics team include data engineers who build and maintain our data architecture and security and monitor data integrity to ensure accuracy in the products created by our data analysts and data scientists. In addition, our data analysts design, build, and manage reports that can be automated; some of them also serve as consultants with HR professionals to help with data storytelling and clarify business needs for our projects.
Littal: Are specific skills or knowledge domains needed to be a data scientist in the HR department as opposed to other departments?
Liz: I believe that understanding ethical and legal standards regarding the use of employee data is incredibly beneficial to a data scientist working in HR. Our People Analytics team works closely with the Legal team to comply with laws regarding the use of employee data. We also have a Privacy team that reviews the use of employee data in reports and predictive models; they make recommendations based on our company standards regarding the protection of and appropriate use of employee data.
Littal: Describe the tools that you use. Are there any R libraries or other open-source tools that are a must? Which is your favorite and why?
Liz: R and Python are open-source tools used by data scientists on our team. I prefer R with RStudio integrated development environment. I have over a dozen R libraries for different projects; my favorite is dplyr – it’s the one I rely on most for data wrangling and probably appears in my R code for every project. For time series forecasting, I use the feasts and fable packages. For text analysis, I use tm and topicmodels packages. The ggplot2 package is excellent for creating beautiful data visualizations.
The contribution of a data scientist to the HR department
Littal: How do business questions related to people are raised? Who initiates them? How do they become a project or a product?
Liz: Our HR Business Partners usually learn about questions from their clients in the Business Units. The HRBP contacts a People Analytics consultant on our team to discuss the project. The consultant then determines which team members should be included in the project and sets up a meeting with the appropriate stakeholders to start defining the project scope and the expectations regarding the project’s final product. The project team continues developing the solution/product and meeting with the client regularly to keep them informed of progress and, if necessary, make changes until the project is complete. If it’s a data science project, the data scientist will be involved in presenting the final product to the business unit leader and HRBP.
Some projects are initiated within our People Analytics team. For example, if we see common themes for requests from multiple business units or persistent requests, we may build a product to satisfy that business needs regularly.
Littal: What are some Ethics considerations in your day-to-day work? Specifically, how do you consider explainability when creating predictive models? Can you share some examples?
Liz: I’ll share a hypothetical example to help explain the ethical considerations. Let’s say a company develops a predictive model determining which employees are likely to leave. If they used a “black box” model that doesn’t allow one to explain the impact of the variables in that model, the result is a model that predicts which employees are likely to leave but doesn’t tell anyone why they are more likely to leave. With this type of model, it would be hard to recommend actions that the company can take to retain valuable employees.
Also, consider how predictive models are used. If misused, a model like this could negatively impact employees and the company. For example, what would happen if a supervisor had access to that information and decided not to assign a team member to a new project because the predictive model indicated that the employee was likely to leave? Perhaps assigning them to a new project will give them a new challenge and eliminate their reason for leaving.
Now consider a situation where a company used an explainable model and was able to understand that employees who have been in the same role with no job growth for years are more likely to leave. Now, the company has some information that can be used to reduce attrition! Perhaps the solution is to help employees learn about available career paths, company training, or other growth opportunities. Again, using explainable models and using them well can be very powerful for HR.
Littal: What are, in your opinion, the fundamentals in data science that should be a part of every HR professional training so that they can be better inner clients of data professionals?
Liz: I think HR professionals should have an introduction to data science with some examples to help them understand what’s possible and how data science can help answer business questions and solve problems. The introduction should include a basic understanding of the different types of predictive models available with HR-related examples of how they can be used. In addition, if the company uses surveys that include employee comments, understanding what’s possible with text analytics can help. If HR professionals have that understanding, they are more likely to recognize a situation where data science can be used, and they can tell their clients, “We can help you with that.” Also, if they understand the process involved in developing data science solutions, it could help set expectations regarding timelines and deliverables early in the conversation with business unit leaders.
Littal: Thank you, Liz, for your insightful perspective. I look forward to finding out how implementing it will lead to more advanced analytics practices within HR departments.