Workforce Analytics Project

📌 Overview

This project investigates employee attrition and workforce segmentation using a combination of statistical analysis, predictive modelling, and clustering. The goal is to help organisations understand who is leaving, who is leading, and what interventions can be taken to improve retention and engagement.

The work draws on a dataset of ~9,000 employees (2023 snapshot), combining demographic, career, and engagement features to:

1. Identify key drivers of voluntary attrition.
2. Segment employees into actionable clusters for targeted talent strategies.

---

🗂 Project Structure

• latest_employee_records.xlsx → Cleaned dataset with one record per employee.
• Technical Document.docx → Methodology, EDA, statistical testing, and modelling details.
• Workforce Analytics (PDF) → Final report with findings, insights, and recommendations.
• Visualisations & Models → Attrition trends, logistic regression outputs, clustering (K-modes, t-SNE).
• Code Notebooks (Python) → Scripts for preprocessing, EDA, statistical testing, logistic regression, and clustering.

---

🔑 Key Use Cases

1. Voluntary Attrition Analysis

• Problem: Identify why employees resign voluntarily.
• Approach:
  • Compared characteristics of leavers vs stayers.
  • Applied t-tests, chi-square, correlation, logistic regression, and random forest.
• Findings:
  • Leavers have fewer promotions (1.62 vs 2.47 for stayers).
  • Attrition risk peaks in 1–3 years tenure and dips after 10+ years.
  • Managers & ERG (Employee Resource Group) members show 60–70% lower quit risk.
• Recommendations:
  • Build structured career pathways.
  • Strengthen ERG membership & visibility.
  • Treat managers as retention anchors.

---

2. Workforce Segmentation (Clustering)

• Problem: Understand natural employee segments for personalised talent strategies.

• Approach:

  • K-modes clustering on categorical features.
  • Chose 7 clusters after elbow method and validation.

• Employee Segments Identified:

  • Early Career Explorers – high attrition risk, low tenure.
  • Consistent Managers – stable, 5–10 yrs tenure, 3–4 promotions.
  • Legacy Leaders – longest-tenured, highly promoted, influential.
  • Cultural Stewards – engaged in BRGs, drive culture.
  • Emerging Talent – top young employees, high flight risk.
  • Senior Talent Hires – new senior recruits adapting to culture.
  • Emerging Leaders – high-potential managers, key to future leadership pipeline.

• Why Prioritise Emerging Leaders?

  • 3–5 years tenure, already managers with 3–4 promotions.
  • At risk of plateauing, but high ROI if engaged.
  • Attrition drops significantly after 3+ promotions (protective factor).

• Recommendations:

  • Offer strategic growth pathways (cross-functional rotations, consulting sprints).
  • Accelerate leadership readiness with curated academies.
  • Recognise & increase visibility in org-wide forums.

---

🛠️ Methodology

1. Data Processing

   • Collapsed monthly HR snapshots into one row per employee.
   • Filtered out involuntary terminations.
   • Converted features into categorical bands (tenure, age, promotions).

2. Exploratory Data Analysis (EDA)

   • Attrition rates by age, tenure, promotions, commute distance, ERG membership, manager role.
   • Dual bar plots for headcount vs exits.

3. Statistical Testing

   • t-tests (continuous variables) and chi-square (categorical variables).
   • Correlation analysis for practical significance.

4. Modelling

   • Logistic Regression: odds ratios for attrition drivers.
   • Random Forest: non-linear interactions and feature importance.

5. Clustering

   • K-modes with Huang initialisation.
   • Optimal k = 7 clusters.
   • t-SNE visualisation for interpretability.

---

🚀 How to Use

1. Dataset: Load latest_employee_records.xlsx.
2. Preprocessing: Run data cleaning pipeline (Python script).
3. Attrition Analysis: Execute statistical tests and models (logistic, random forest).
4. Clustering: Run K-modes script to generate 7 clusters.
5. Visualisation: Review charts for attrition trends and cluster profiles.
6. Interpret Results: Refer to the final report (Workforce Analytics.pdf) for insights and recommendations.

---

📊 Tools & Technologies

• Languages: Python (Pandas, NumPy, Matplotlib, Seaborn, Scikit-learn).
• Statistical Methods: T-test, Chi-square, Correlation, Logistic Regression, Random Forest.
• Clustering: K-modes, t-SNE.
• Visualisation: Tableau & Matplotlib.
• Data Source: HRIS employee records (2023 snapshot).

---

📌 Future Work

• Expand analysis to include environmental & external labour market data.
• Track involuntary attrition separately.
• Deploy interactive dashboards for HR leaders.
• Extend segmentation to other regions & survey data.