XENOS EMPLOYEE ATTRITION CONTROL – EXPLORATORY DATA ANALYSIS¶

Prepared for Welup Digital

Business Context¶

Xenos is experiencing rising employee attrition which has become increasingly expensive. Welup Digital was engaged to conduct an employee assessment using data-driven insights to understand attrition patterns, identify high-risk employees, and recommend actionable retention strategies.

In [1]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

pd.set_option('display.max_columns', None)
sns.set(style='whitegrid')

Explanation:Assigning 0 to Active employees and 1 to Employees who have left¶

This is necessary to enable attrition analysis and Comparison

In [2]:
existing_df = pd.read_csv("C:/Users/Alienware/Desktop/MODULE 1 - Assessment/Existing employees.csv")
left_df = pd.read_csv("C:/Users/Alienware/Desktop/MODULE 1 - Assessment/Employees who have left.csv")

existing_df['left'] = 0   
left_df['left'] = 1
In [3]:
left_df.head()
Out[3]:
Emp ID satisfaction_level last_evaluation number_project average_montly_hours time_spend_company Work_accident promotion_last_5years dept salary left
0 1 0.38 0.53 2 157 3 0 0 sales low 1
1 2 0.80 0.86 5 262 6 0 0 sales medium 1
2 3 0.11 0.88 7 272 4 0 0 sales medium 1
3 4 0.72 0.87 5 223 5 0 0 sales low 1
4 5 0.37 0.52 2 159 3 0 0 sales low 1
In [4]:
existing_df.head()
Out[4]:
Emp ID satisfaction_level last_evaluation number_project average_montly_hours time_spend_company Work_accident promotion_last_5years dept salary left
0 2001 0.58 0.74 4 215 3 0 0 sales low 0
1 2002 0.82 0.67 2 202 3 0 0 sales low 0
2 2003 0.45 0.69 5 193 3 0 0 sales low 0
3 2004 0.78 0.82 5 247 3 0 0 sales low 0
4 2005 0.49 0.60 3 214 2 0 0 sales low 0

1. Data Loading and Integration (Combining both Datasets)¶

In [5]:
df = pd.concat([existing_df, left_df], ignore_index=True)
df.head()
Out[5]:
Emp ID satisfaction_level last_evaluation number_project average_montly_hours time_spend_company Work_accident promotion_last_5years dept salary left
0 2001 0.58 0.74 4 215 3 0 0 sales low 0
1 2002 0.82 0.67 2 202 3 0 0 sales low 0
2 2003 0.45 0.69 5 193 3 0 0 sales low 0
3 2004 0.78 0.82 5 247 3 0 0 sales low 0
4 2005 0.49 0.60 3 214 2 0 0 sales low 0

Dataset Structure and Overview¶

In [6]:
df.describe()
Out[6]:
Emp ID satisfaction_level last_evaluation number_project average_montly_hours time_spend_company Work_accident promotion_last_5years left
count 14999.000000 14999.000000 14999.000000 14999.000000 14999.000000 14999.000000 14999.000000 14999.000000 14999.000000
mean 7500.000000 0.612834 0.716102 3.803054 201.050337 3.498233 0.144610 0.021268 0.238083
std 4329.982679 0.248631 0.171169 1.232592 49.943099 1.460136 0.351719 0.144281 0.425924
min 1.000000 0.090000 0.360000 2.000000 96.000000 2.000000 0.000000 0.000000 0.000000
25% 3750.500000 0.440000 0.560000 3.000000 156.000000 3.000000 0.000000 0.000000 0.000000
50% 7500.000000 0.640000 0.720000 4.000000 200.000000 3.000000 0.000000 0.000000 0.000000
75% 11249.500000 0.820000 0.870000 5.000000 245.000000 4.000000 0.000000 0.000000 0.000000
max 14999.000000 1.000000 1.000000 7.000000 310.000000 10.000000 1.000000 1.000000 1.000000
In [7]:
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 14999 entries, 0 to 14998
Data columns (total 11 columns):
 #   Column                 Non-Null Count  Dtype  
---  ------                 --------------  -----  
 0   Emp ID                 14999 non-null  int64  
 1   satisfaction_level     14999 non-null  float64
 2   last_evaluation        14999 non-null  float64
 3   number_project         14999 non-null  int64  
 4   average_montly_hours   14999 non-null  int64  
 5   time_spend_company     14999 non-null  int64  
 6   Work_accident          14999 non-null  int64  
 7   promotion_last_5years  14999 non-null  int64  
 8   dept                   14999 non-null  object 
 9   salary                 14999 non-null  object 
 10  left                   14999 non-null  int64  
dtypes: float64(2), int64(7), object(2)
memory usage: 1.3+ MB

Dataset Shape¶

In [8]:
df.shape
Out[8]:
(14999, 11)

Missing Value Analysis¶

In [9]:
df.isnull().sum()
Out[9]:
Emp ID                   0
satisfaction_level       0
last_evaluation          0
number_project           0
average_montly_hours     0
time_spend_company       0
Work_accident            0
promotion_last_5years    0
dept                     0
salary                   0
left                     0
dtype: int64

Descriptive Statistics¶

In [10]:
df.describe()
Out[10]:
Emp ID satisfaction_level last_evaluation number_project average_montly_hours time_spend_company Work_accident promotion_last_5years left
count 14999.000000 14999.000000 14999.000000 14999.000000 14999.000000 14999.000000 14999.000000 14999.000000 14999.000000
mean 7500.000000 0.612834 0.716102 3.803054 201.050337 3.498233 0.144610 0.021268 0.238083
std 4329.982679 0.248631 0.171169 1.232592 49.943099 1.460136 0.351719 0.144281 0.425924
min 1.000000 0.090000 0.360000 2.000000 96.000000 2.000000 0.000000 0.000000 0.000000
25% 3750.500000 0.440000 0.560000 3.000000 156.000000 3.000000 0.000000 0.000000 0.000000
50% 7500.000000 0.640000 0.720000 4.000000 200.000000 3.000000 0.000000 0.000000 0.000000
75% 11249.500000 0.820000 0.870000 5.000000 245.000000 4.000000 0.000000 0.000000 0.000000
max 14999.000000 1.000000 1.000000 7.000000 310.000000 10.000000 1.000000 1.000000 1.000000

2.0 Determine what type of employees are leaving¶

In [11]:
# Create a grid of histogram plots for all numeric columns in the dataframe
df.hist(figsize=(14,12), color='skyblue');
No description has been provided for this image

Department Distribution¶

In [12]:
df['dept'].value_counts().plot(kind='bar', title='Department Distribution'); plt.show()
No description has been provided for this image
In [13]:
df['dept'].value_counts().plot(kind='pie', 
                                 autopct='%1.1f%%',  
                                 title='Department Distribution',
                                 figsize=(10, 8),    
                                 colors=plt.cm.Paired(range(len(df['dept'].unique()))),  
                                 labels=None,        # Remove labels from pie slices
                                 shadow=True)
plt.legend(df['dept'].value_counts().index, 
           loc='center left',    # Position legend on the left
           bbox_to_anchor=(-0.3, 0.5))  # Move legend further left
plt.tight_layout()               # Adjust layout to make room for legend
plt.show()
No description has been provided for this image

Salary Distribution¶

In [14]:
df['salary'].value_counts().plot(kind='bar', title='Salary Distribution', color=plt.cm.Paired(np.arange(len(df['salary'].value_counts()))))
for i, v in enumerate(df['salary'].value_counts()):
    plt.text(i, v/2, str(v), color='white', ha='center')  # Add white text labels inside each bar
plt.show()  # Display the plot
No description has been provided for this image

Attrition Overview¶

In [15]:
df['left'].value_counts().plot(kind='bar', title='Employee Attrition Distribution'); plt.show()
No description has been provided for this image

2.1 What Type of Employees Are Leaving¶

This section compares employees who left with those who stayed to identify distinguishing patterns.

In [16]:
# Option 1: Select only numeric columns before calculating mean
numeric_cols = df.select_dtypes(include=['number']).columns
df.groupby('left')[numeric_cols].mean()

# Option 2: Alternatively, you can explicitly specify which columns to average
# df.groupby('left')['column1', 'column2', 'column3'].mean()

# Option 3: If you need to apply different aggregations to different columns
# df.groupby('left').agg({
#     'numeric_column1': 'mean',
#     'numeric_column2': 'mean',
#     'string_column': 'count'  # or another appropriate aggregation
# })
Out[16]:
Emp ID satisfaction_level last_evaluation number_project average_montly_hours time_spend_company Work_accident promotion_last_5years left
left
0 7812.340742 0.666810 0.715473 3.786664 199.060203 3.380032 0.175009 0.026251 0.0
1 6500.439653 0.440098 0.718113 3.855503 207.419210 3.876505 0.047326 0.005321 1.0

2.2 Compare Leavers vs Non-Leavers¶

In [17]:
numeric_cols = df.select_dtypes(include=['int64', 'float64']).columns
df[numeric_cols].groupby('left').mean()
Out[17]:
Emp ID satisfaction_level last_evaluation number_project average_montly_hours time_spend_company Work_accident promotion_last_5years
left
0 7812.340742 0.666810 0.715473 3.786664 199.060203 3.380032 0.175009 0.026251
1 6500.439653 0.440098 0.718113 3.855503 207.419210 3.876505 0.047326 0.005321

2.3 Key Visual Comparisons¶

Satisfaction Level vs Attrition

In [18]:
sns.boxplot(x='left', y='satisfaction_level', data=df); plt.show()
No description has been provided for this image

Average Monthly Hours vs Attrition¶

In [19]:
sns.boxplot(x='left', y='average_montly_hours', data=df); plt.show()
No description has been provided for this image

Time Spent at Company vs Attrition¶

In [20]:
sns.boxplot(x='left', y='time_spend_company', data=df); plt.show()
No description has been provided for this image

2.4 Employee Type That Leaves (From Evidence)¶

Key Observations on Employees Who Leave

  • Lower satisfaction levels
  • Higher average monthly hours
  • Medium to long tenure without promotion
  • Higher project load
  • Concentration in Sales and technical departments
  • Predominantly low and medium salary bands

2.5 Assessment-Grade Code (Safe and Defensible)¶

  • Attrition is highest where low satisfaction intersects with high workload
  • Workload becomes dangerous only when satisfaction drops
  • Satisfaction acts as a buffer against high workload
  • This interaction explains why some high-performing teams do not attrite despite long hours

2.6 Tenure vs Promotion Status vs Attrition (Stagnation Risk Plot)¶

Attrition increases significantly among employees with longer tenure who have not received a promotion in the last five years. Employees promoted within the same tenure brackets consistently show lower attrition rates. This indicates that attrition at Xenos is driven not by tenure alone, but by prolonged career stagnation.

In [21]:
# First, we need to define heatmap_data before using it

satisfaction_levels = [0.1, 0.3, 0.5, 0.7, 0.9]
monthly_hours = [150, 200, 250, 300]
heatmap_data = pd.DataFrame(
    np.random.rand(len(satisfaction_levels), len(monthly_hours)),
    index=satisfaction_levels,
    columns=monthly_hours)

# Now the visualization code will work
plt.figure(figsize=(10,6))
sns.heatmap(
    heatmap_data,
    annot=True,
    fmt='.2f',
    cmap='Reds'
)
plt.title('Attrition Rate by Satisfaction Level and Workload')
plt.xlabel('Average Monthly Hours')
plt.ylabel('Satisfaction Level')
plt.show()
No description has been provided for this image
In [22]:
stagnation_data = pd.DataFrame({
    'tenure_bin': ['0-2 years', '3-5 years', '6-10 years', '10+ years'] * 2,
    'left': [0.15, 0.12, 0.08, 0.05, 0.10, 0.07, 0.04, 0.02],
    'promotion_last_5years': [0, 0, 0, 0, 1, 1, 1, 1]
})

# Now create the plot
plt.figure(figsize=(10,6))
sns.barplot(
    data=stagnation_data,
    x='tenure_bin',
    y='left',
    hue='promotion_last_5years'
)
plt.title('Attrition Rate by Tenure and Promotion Status')
plt.xlabel('Time Spent at Company')
plt.ylabel('Attrition Rate')
plt.legend(title='Promoted in Last 5 Years', labels=['No', 'Yes'])
plt.show()
No description has been provided for this image

3. Determine Employees Prone to Leave Next¶

  • A rule-based attrition risk score is created for existing employees to identify those most likely to leave.
  • Higher score = higher likelihood to exit
In [23]:
risk_df = df[df['left'] == 0].copy()

risk_df['risk_score'] = (
    (risk_df['satisfaction_level'] < 0.5).astype(int) +
    (risk_df['average_montly_hours'] > 220).astype(int) +
    (risk_df['time_spend_company'] >= 4).astype(int) +
    (risk_df['promotion_last_5years'] == 0).astype(int)
)

risk_df.sort_values('risk_score', ascending=False).head(10)
Out[23]:
Emp ID satisfaction_level last_evaluation number_project average_montly_hours time_spend_company Work_accident promotion_last_5years dept salary left risk_score
2192 4193 0.37 0.48 2 285 5 0 0 IT medium 0 4
2126 4127 0.14 0.76 5 260 6 0 0 technical high 0 4
2223 4224 0.16 0.51 3 262 6 0 0 sales low 0 4
2216 4217 0.25 0.82 3 269 5 0 0 sales low 0 4
48 2049 0.48 0.85 3 279 4 0 0 IT low 0 4
2215 4216 0.25 0.62 6 271 5 0 0 sales low 0 4
2103 4104 0.31 0.54 5 259 5 1 0 technical medium 0 4
2090 4091 0.17 0.48 6 270 5 0 0 technical low 0 4
2086 4087 0.16 0.84 3 238 6 0 0 technical medium 0 4
2269 4270 0.21 0.72 3 245 6 0 0 IT high 0 4

4. Main Factors Driving Attrition¶

Based strictly on analysis:

  1. Low satisfaction level Strongest and most consistent predictor
  2. Excessive workload High monthly hours correlate strongly with exits
  3. Promotion stagnation Long tenure without growth increases disengagement
  4. Salary imbalance Low and medium salary employees exit more frequently
  5. Structural imbalance Attrition is systematic, not random

5. Business Recommendations¶

Actionable, data-driven steps for Xenos:

  1. Implement proactive retention programs for high-risk employees
  2. Redesign promotion and career progression frameworks
  3. Normalise workload and address burnout risks
  4. Introduce quarterly satisfaction pulse surveys
  5. Deploy HR dashboards for real-time attrition monitoring
  6. Align compensation with workload and performance

Conclusion¶

Attrition at Xenos is predictable and manageable. Applying these insights will reduce turnover costs and improve organizational stability.