Stay or Go? Predicting Employee Turnover with HR Analytics

Not long ago, I was chatting with a friend in HR who was stressed about keeping younger employees from leaving. That stuck with me. What if data could help answer why people leave... and who’s most likely to stay? That conversation kicked off this project, where I dove into HR data to look for patterns behind employee turnover and income.

Why This Project?

The idea for this project began with a desire to apply analytics to something impactful. I was intrigued by the challenge of employee turnover and how HR data could shed light on it. I gathered information on tenure, role changes, performance ratings, and exit interviews to explore why some employees stay and others leave. This project was more than just analyzing spreadsheets—it was about understanding the human stories behind retention metrics. I aimed to address a significant business question:

What factors most accurately predict employee turnover, and how can organizations use these insights to improve retention?

What You Will Gain:

• What the data says about employee attrition (who leaves and why)

• What really drives employee income (it’s not what you might think)

• Surprising lessons from the data that challenged my assumptions

Key Insights:

• Younger employees are more likely to leave than older ones.

• Age alone doesn’t explain much about someone’s salary.

• Job level and total experience are much stronger indicators of income.

Dataset Details:

The dataset comes from Kaggle and includes details on 1,470 employees; things like age, education, job role, salary, and whether they left the company. It was well-organized and had 35 columns, making it a great fit for analysis and modeling in R.

Analysis Process:

First, I explored the data and cleaned it where needed. Then I ran statistical tests (like a t-test) to see if there were real differences in age between those who left and those who stayed. I followed that with simple and multivariable regression models to predict monthly income.

One of the biggest surprises? Age looked important at first. But once I added job level and total working years to the model, age lost its predictive power. That was a reminder that the most obvious answer isn’t always the most accurate and that context really matters.

Visuals and Insights:

   To get a high-level view of what drives employee attrition, this correlation matrix reveals how strongly each variable is related to the likelihood of someone leaving. Factors like age, income, and tenure show particularly strong relationships, helping us zero in on the most influential drivers of turnover.

  To uncover which factors are most associated with employee attrition, this multi-panel plot shows the relationships between attrition and variables like income, age, tenure, and education. Each box reveals how patterns differ between those who stayed and those who left, with correlation values highlighting where the strongest links lie—particularly around total working years, age, and income. These patterns help us focus on the drivers most predictive of turnover.

To see if younger employees are more likely to leave, this chart compares the average age of those who stayed versus those who left. Employees who left were, on average, younger; and the difference is statistically meaningful, as shown by the non-overlapping confidence intervals.

To find out if lower-level employees are more likely to leave, we looked at income across job levels. The results are clear: those in Job Levels 1 and 2 earn significantly less, with far less variation in pay. This steep income gap may be more than just a number. It could be a key driver behind higher turnover among junior staff.

Main Takeaways:

Looking at the data holistically helped me see the full picture. While age initially seemed important, it wasn’t the best predictor when you factored in experience and job level. The real takeaway is that employees earn more because they’ve worked longer or advanced further, not simply because they’re older.

This project showed me how valuable it is to go beyond surface-level patterns and combine different tools...like visualizations, stats, and models to find deeper truths in the data.

Conclusion and Personal Reflections:

This project taught me how to think critically about data, test assumptions, and communicate findings in a way that makes sense to others. It challenged me to balance technical depth with clarity, and it reminded me that the best insights are often the ones that make you rethink your first impression.

Call To Action:

If this project sparked your interest or you're working on something similar, I’d love to connect on LinkedIn. And if you know someone looking for a data analyst who’s passionate about solving real problems with data...let’s talk!