Capstone 3 — HR Attrition Analytics¶
A self-directed capstone analysing why employees leave and what an organisation can do about it. This capstone also tests your ethical judgment — HR data is about real people.
The Brief¶
A company's attrition is rising and costing millions. HR wants to understand the drivers and where to focus retention efforts. You analyse the data, identify actionable drivers, and recommend interventions — while respecting employee privacy and avoiding discriminatory conclusions.
Tools¶
Python (Pandas) · SQL · Power BI/Tableau
Difficulty¶
Beginner-Intermediate — clean data, clear target. The challenge is in the ethical framing and actionability.
What You Must Deliver¶
- A clean HR dataset with engineered features
- An attrition analysis across multiple dimensions
- Identification of actionable vs non-actionable drivers
- A privacy-respecting dashboard (aggregate, not individual surveillance)
- Ethical, actionable recommendations
Suggested Datasets¶
- IBM HR Analytics Employee Attrition (Kaggle) — the standard
- Any HR dataset with an attrition/turnover flag
Business Questions¶
- What is the overall attrition rate and its estimated cost?
- Which departments/roles have the highest attrition?
- What factors correlate most with leaving?
- Which of those factors can the company actually change?
- Where should a retention budget be focused?
Requirements Checklist¶
- [ ] Attrition analysed by department, role, tenure, and key factors
- [ ] Correlation analysis identifying the strongest drivers
- [ ] Clear separation of actionable (overtime, pay, development) vs non-actionable (age, marital status) factors
- [ ] Privacy-respecting dashboard (minimum group sizes, no individual flagging)
- [ ] Ethical recommendations focused on systemic improvement
- [ ] Portfolio README
The Ethical Requirement¶
This capstone tests ethical judgment
You must explicitly identify which correlations you will NOT act on (e.g., marital status, age — acting on these is discriminatory and often illegal). Strong submissions focus exclusively on factors the company can ethically change, and avoid building individual "flight risk" surveillance. See Projects/HR-Analytics-Dashboard/interview-questions for the reasoning.
Evaluation Criteria¶
| Criterion | Weight |
|---|---|
| Analysis across dimensions | 20% |
| Driver identification | 20% |
| Ethical judgment | 25% |
| Actionability of recommendations | 20% |
| Dashboard & communication | 15% |
Stretch Goals¶
- Build a logistic regression to identify systemic risk factors (not individual surveillance)
- Estimate the ROI of a specific retention intervention
- Manager-level analysis ("people leave managers")