📊 HR Attrition Analysis | IBM

Analysed by evanalytics

This project simulates how a Financial Analyst can leverage HR attrition data to support forecasting, cost modeling, and retention strategy — all essential to modern FP&A functions.

---

🎯 Objective

To explore how data-driven HR insights can inform financial decisions, we:

• Model the cost of employee attrition
• Detect potential retention bottlenecks
• Simulate real-world FP&A use cases

---

🧪 Hypotheses

1. Low income, frequent overtime, and low satisfaction are predictors of attrition.
2. Departments with high attrition also show high variance in tenure and performance.
3. Younger employees tend to leave earlier, reducing ROI on onboarding.

---

🛠️ Tech Stack

Category	Tools
Analysis	Python (Pandas, Seaborn, Matplotlib, Plotly)
Interface	Jupyter Notebook
Automation	Scripts & GitHub
IDE	VS Code with Git Bash
Visualization	Tableau / Power BI (optional)

---

🗂️ Project Structure

HR_analytics_evCorp/
.
├── data
│   └── WA_Fn-UseC_-HR-Employee-Attrition.csv
├── kaggle.json
├── notebooks
│   ├── correlation_heatmap.ipynb
│   ├── explore_dataset.ipynb
│   ├── explore_dataset2.ipynb
│   └── plot_attrition_cost.ipynb
├── outputs
│   ├── attrition_cost_by_department.csv
│   ├── attrition_leavers.csv
│   ├── correlation_heatmap.png
│   ├── percentages_attrition_rate_by_department.png
│   ├── plot_attrition_cost.png
│   ├── plot_Number_of_Employees_Who_Left.png
│   └── share_of_total_attrition_by_department.png
├── README.md
├── requirements.txt
├── scripts
│   ├── analyze_hr_data.py
│   ├── load_kaggle_data.py
│   └── model_attrition_cost.py
├── structure.txt
├── venv
│   ├── bin
│   ├── etc
│   ├── include
│   ├── lib
│   ├── pyvenv.cfg
│   └── share
└── venv310
    ├── etc
    ├── Lib
    ├── pyvenv.cfg
    ├── Scripts
    └── share

16 directories, 21 files

---

🔍 Attrition Cost Modeling

The script scripts/model_attrition_cost.py loads the dataset and calculates financial impact using the following steps:

What it does:

1. Loads data from /data/WA_Fn-UseC_-HR-Employee-Attrition.csv

2. Flags records where Attrition == 'Yes'

3. Assigns a fixed cost of $50,000 per leaver (editable in script)

4. Computes:

   • 🔻 Total number of leavers
   • 💸 Total estimated cost of attrition
   • 🏢 Department-wise cost breakdown

5. Saves results to /outputs folder as CSV files

6. Plots a horizontal bar chart showing departmental attrition cost

Output Files:

File	Description
outputs/attrition_leavers.csv	Records of employees who left
outputs/attrition_cost_by_department.csv	Cost breakdown per department

---

📊 Sample Insights

---

📊 Attrition Cost

🧠 Attrition by Job Satisfaction Level

This chart shows the attrition rate grouped by employee job satisfaction.

• Low satisfaction = highest attrition
• Very High satisfaction = lowest attrition
• The trend is clear and visual: satisfied employees stay longer

---

🏢 Attrition by Department

Departments with higher headcount loss represent greater risk in operational continuity and increased hiring/training costs.

---

👥 Attrition by Job Role

Some job roles — especially client-facing or high-stress roles — exhibit notably higher attrition rates. This helps target role-specific retention programs.

👥 Attrition by Job Satisfaction

👥 Attrition by Overtime

---

👥 Total Leavers by Department

---

• Correlation heatmaps revealed patterns in satisfaction, income, and overtime
• Attrition rates were highest in Sales and R&D
• Simulated savings by improving retention in high-cost departments

Explanation of Correlation Heatmap

---

📊 What You’re Seeing

This is a correlation matrix heatmap, where:

• Each square represents the correlation between two variables.

• Values range from:

  • +1.00: Perfect positive correlation (when one goes up, so does the other)
  • 0.00: No correlation
  • -1.00: Perfect negative correlation (when one goes up, the other goes down)

• Colors:

  • 🔴 Redder = stronger correlation (positive or negative)
  • 🔵 Bluer = weaker or no correlation

---

🔎 Focus: Correlation with Attrition

You want to understand what variables are related to employee attrition — so let's read across the Attrition row:

Variable	Correlation	Meaning
OverTime	+0.25	Strongest positive — more overtime → higher chance of attrition ✅
MonthlyIncome	-0.16	Negative — lower income → higher chance of leaving
JobLevel	-0.17	Lower job level → more likely to leave
TotalWorkingYears	-0.17	Fewer working years → more likely to leave
Age	-0.16	Younger employees are more likely to leave
YearsAtCompany	-0.13	Less time at company → higher attrition risk

Everything else is near 0.00, so weak or no relationship.

✅ Most Insightful Patterns

1. OverTime is the most powerful predictor

• Correlation = +0.25
• Suggests employees working overtime are more likely to quit (burnout risk)

2. Lower MonthlyIncome, Lower JobLevel, Younger Age

• All have negative correlations with attrition

• Suggests people leave when they:

  • Feel underpaid
  • Are in junior roles
  • Are early in their career

3. Tenure indicators matter

• YearsAtCompany, TotalWorkingYears: negative correlation
• Less experience → more likely to leave (possibly due to unmet expectations or lack of fit)

⚠️ Caution

• These are correlations, not causation.
• Just because OverTime is positively correlated with Attrition doesn’t prove it causes attrition, but it's a strong signal.

💡 Business Value

• Helps HR & Finance align on cost-saving goals
• Aids in budgeting and strategic workforce planning
• Demonstrates how analytics supports cost control