Project Background

This project aims to analyze and predict employee attrition using a fictional dataset created by IBM data scientists. This dataset contains various features that may influence employee turnover, such as job roles, monthly income, education, and distance from home. By conducting exploratory data analysis (EDA), we aim to uncover key factors contributing to attrition, including insights like job role distribution and how distance from home correlates with attrition.

After the data cleaning process, which involves handling missing values, outliers, and inconsistent data entries, we proceed with feature engineering and visualization to better understand the dataset. The ultimate objective is to build and evaluate a predictive model using classification algorithms to determine whether an employee will likely leave the company. HR teams can leverage this to create targeted retention strategies and improve employee satisfaction.

Key Insights

• Top Influencers of Attrition:

  1. Monthly Income
  2. Age
  3. Total Working Years
  4. Years at Company
  5. Overtime (Yes)

• Early-Career Professional Cluster:

  • Employees in this category, clustered by MonthlyIncome, TotalWorkingYears, and YearsAtCompany, exhibit the highest churn rates, with a retention rate of just 3:1.

• Attrition and Age:

  • Younger employees, are most likely to leave, closely aligning with the early-career group averaging 31.14 years for males and 30.61 for females.

• Income Disparities:

  • Early-career professionals earn an average of 3459.08, significantly below the organisational average of 6502.93.

• Job Level and Retention:

  • Attrition is notably higher in entry-level roles. Job Level 1 employees have a retention rate of 3:1, with an average income of 2786.92, while Job Level 5 sees a retention rate of 13:1, with average earnings of 19191.83.

• Overtime and Attrition Impact:

  • Overtime work is consistent across clusters, with only 4.5% rating their work-life balance poorly. This indicates that attrition is driven by a combination of factors, where financial compensation alongside overtime plays a significant role.

The Jupyter Notebook used to inspect and clean the data for this analysis and EDA can be found here link.

An interactive Power BI dashboard used to report and explore sales trends can be found here link.

Data Structure & Initial Checks

The datacard describiing this dataset can be found at Kaggle A short summary of the features is described below:

Feature	Data Type	Description
Age	Integer	The age of the employee.
Attrition	String	Whether the employee has left the organisation or not.
BusinessTravel	String	Frequency which an employee travels.
DailyRate	Integer	Daily rate of the employee.
Department	String	Name of the department the employee works in.
DistanceFromHome	Integer	Distance the employee has to travel to work.
Education	Integer	The education level of the employee.
EducationField	String	What the employee studied during their education.
EmployeeCOunt	Integer	A binary count of all the employees in the organisation.
EmployeeNumber	ID	Identification number for each employee.
EnvironmentalSatisfaction	Integer	Satisfaction rating of the work enviornment on a scale of 1 to 4.
Gender	String	The gender of the employee.
HourlyRate	Integer	The employees rate per hour.
JobInvolvement	Integer	Involvement rating of the work required for role, on a scale of 1 to 4.
JobLevel	Integer	The hierachial level of the job responsibilities of the employee, from a scale of 1 to 4.
JobRole	String	The job the employee does within the organisation.
JobSatisfaction	Integer	The level of satisfaction in the work the employee does in their role on a scale of 1 to 4.
MaritalStatus	String	The employee's marital status.
MonthlyIncome	Integer	The monthly income of the employee.
MonthlyRate	Integer	The rate of the employee per month.
NumOfCompaniesWorked	Integer	The total number of companies the employee has worked for before.
Over18	String	Whether the employee is over the age of 18 or not.
OverTime	String	Whether the employee has worked overtime or not.
PercentSalaryHike	Integer	The percentage increase an employee received from their previous monthly income.
PerformanceRating	Integer	The rating of the employees performance within the organisation, on a scale of 1 to 4.
RelationshipSatisfaction	Integer	The rating of the relationship the employee has with other staff memebers within the organisation.
StadardHours	Integer	The standard working hours of the employee.
StockOptionLevel	Integer	Whether the employee has stock level options, and what level the stick level option is. Value greater than 0 is represents stock option for the employee.
TotalWorkingYears	Integer	The total working years for the employee.
TrainingTimesLastYear	Integer	The total number of times the employee received training the previous year.
WorkLifeBalance	Integer	A rating of the work-life balance of the employee from a scale of 1 to 4.
YearsAtCompany	Integer	The total number of years the employee has been in the organisation.
YearsInCurrentRole	Integer	The total number of years the employee has been in their current job role.
YearsSinceLastPromotion	Integer	The total number of years since the employees last promotion.
YearsWithCurrManager	Integer	The total number of years spent with their current manager.

Executive Summary

The analysis identifies key drivers of employee attrition, with compensation, career stage, and job level emerging as critical factors. Early-career professionals and entry-level employees show the highest churn rates, driven by lower earnings and limited progression opportunities. Notably, younger employees, especially those in early-career clusters, exhibit a strong tendency toward attrition, with those aged around 30 earning well below the organisational average.

Additionally, while overtime rates are consistent across clusters, their impact on attrition varies based on financial compensation, underscoring the importance of aligning work-life balance initiatives with competitive remuneration.

An overview of this information can be found in the PowerBi Dashboard

Insights Deep Dive

Retention rate per Clusters

The retention rates of each cluster can be found below. Based on the data in the graph, we can see that the employee clusters with the worst retention rates are the Early-career professionals and Veteran Professionals.

Average MonthlyIncome per Employee Clustering

The monthly income of early-career employees is well below that of the organisation's mean income value. Knowing that income is a strong indicator for attrition, it brings further understanding on why early-career employees have the worst retention rate within the organisation.

Total Attrition per Job Level

Based on the barplot, we can see that job roles in job level 1 have the worst retention rate. This further reiterates the inference that early-career professionals are the group of employees most likely to churn. This can be seen in how all the job roles of the early-career professionals are level 1 and level 2 job levels.

Recommendations

Increase Compensation for Early-Career Professionals: Adjust pay for early-career roles to be closer to the organisational average, addressing financial disparity and enhancing retention in this key segment.

Develop Career Progression Pathways: Establish clear development tracks, especially for entry-level roles, to fulfill employees' growth aspirations and improve engagement.

Engagement Programmes for Younger Employees: Introduce mentorship and career coaching tailored to younger talent, addressing their unique needs and improving satisfaction.

Supportive Work-Life Balance Initiatives: Enhance wellness and flexible working options, providing a balanced environment that mitigates attrition even among those working overtime.

Adjust Entry-Level Compensation: Address disparities in pay for lower job levels to boost satisfaction and engagement, reducing turnover at the organisation's entry points.