Call Center Performance Analysis: A Data-Driven Approach to Agent Productivity

1. Identifying Top-Performing Agents

In any call center environment, evaluating the performance of agents is crucial to ensuring high-quality customer service. For this project, I analyzed the dataset to identify the top-performing agents based on their ratings. Agent ratings are an indicator of an agent's ability to handle calls effectively, satisfy customers, and achieve desired call outcomes. Identifying top performers helps in recognizing and rewarding excellence while also setting a standard for others to emulate.

# Import necessary libraries
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import numpy as np

# Load the dataset
file_path = r"C:\Users\user\Documents\PYTHON PROJECTS\Call Center\Data\Sales and Marketing Call Center.csv"
df = pd.read_csv(file_path)

# Create a full name column to enhance identification
df['Agent_Full_Name'] = df['Agent_First_Name'] + ' ' + df['Agent_Last_Name']

# Convert 'Agent_Rating' to numeric for accurate filtering
df['Agent_Rating'] = pd.to_numeric(df['Agent_Rating'], errors='coerce')

# Filter and display top-performing agents with a perfect rating of 5.0
top_agents = df[df['Agent_Rating'] == 5.0][['Agent_Full_Name', 'Agent_Rating']].drop_duplicates()

# Display the result
print("Top-Performing Agents Based on Agent Rating:")
print(top_agents)
      

2. Average Call Duration by Each Agent

# Ensure 'Call_Duration_Minutes' is numeric
df['Call_Duration_Minutes'] = pd.to_numeric(df['Call_Duration_Minutes'], errors='coerce')

# Group by 'Agent_Full_Name' and calculate the average call duration, then sort in descending order
average_call_duration = df.groupby('Agent_Full_Name')['Call_Duration_Minutes'].mean().round(2).reset_index()
average_call_duration.rename(columns={'Call_Duration_Minutes': 'Average_Call_Duration'}, inplace=True)
average_call_duration = average_call_duration.sort_values(by='Average_Call_Duration', ascending=False)

# Print the heading
print(f"{'Agent Full Name':<30} | {'Average Call Duration (minutes)':<10}")
print("-" * 50)

# Display each agent's full name followed by their average call duration with proper spacing
for _, row in average_call_duration.iterrows():
  print(f"{row['Agent_Full_Name']:<30} | {row['Average_Call_Duration']:<10} minutes")
      

3. Agents with the Highest Success Rate in Closing Calls

# Ensure 'Call_Outcome' column is properly labeled and 'Success' is counted
df['Call_Outcome'] = df['Call_Outcome'].str.strip()  # To handle any extra spaces if present

# Filter out the successful calls
success_calls = df[df['Call_Outcome'] == 'Success']

# Group by 'Agent_Full_Name' and calculate the total and successful calls
agent_success = success_calls.groupby('Agent_Full_Name').size().reset_index(name='Total_Successful_Calls')

# Calculate total calls per agent
total_calls = df.groupby('Agent_Full_Name').size().reset_index(name='Total_Calls')

# Merge the dataframes to have total and successful calls together
agent_data = pd.merge(agent_success, total_calls, on='Agent_Full_Name')

# Calculate the success rate as a percentage
agent_data['Success_Rate'] = (agent_data['Total_Successful_Calls'] / agent_data['Total_Calls']) * 100

# Round the success rate to 2 decimal places
agent_data['Success_Rate'] = agent_data['Success_Rate'].round(2)

# Sort the data by Success Rate in descending order
agent_data_sorted = agent_data.sort_values(by='Success_Rate', ascending=False)

# Print the heading
print(f"{'Agent Full Name':<30} | {'Total Successful Calls':<20} | {'Success Rate (%)':<15}")
print("-" * 70)

# Display each agent's full name, total successful calls, and success rate
for _, row in agent_data_sorted.iterrows():
  print(f"{row['Agent_Full_Name']:<30} | {row['Total_Successful_Calls']:<20} | {row['Success_Rate']:<15}")
      

Success Rate Visualization

Evaluating the success rate of call center agents is essential for measuring performance and improving customer service. In this project, I analyzed agent call outcomes, calculating their total successful calls and overall success rates. A visual representation using a bar and line plot highlights the top-performing agents and their efficiency in handling calls.

# Ensure 'Call_Outcome' column is cleaned
df['Call_Outcome'] = df['Call_Outcome'].str.strip()

# Calculate total successful calls
success_calls = df[df['Call_Outcome'] == 'Success']
agent_success = success_calls.groupby('Agent_Full_Name').size().reset_index(name='Total_Successful_Calls')

# Calculate total calls for each agent
total_calls = df.groupby('Agent_Full_Name').size().reset_index(name='Total_Calls')

# Merge and compute success rate
agent_data = pd.merge(agent_success, total_calls, on='Agent_Full_Name')
agent_data['Success_Rate'] = (agent_data['Total_Successful_Calls'] / agent_data['Total_Calls']) * 100
agent_data['Success_Rate'] = agent_data['Success_Rate'].round(2)

# Sort data for visualization
agent_data_sorted = agent_data.sort_values(by='Success_Rate', ascending=False)

# Create the plot
plt.figure(figsize=(14, 8))
sns.set_style("whitegrid")

# Bar plot for successful calls
bar_plot = sns.barplot(
  x='Agent_Full_Name',
  y='Total_Successful_Calls',
  data=agent_data_sorted,
  color='skyblue',
  edgecolor='black'
)

# Line plot for success rate
line_plot = bar_plot.twinx()
line_plot.plot(agent_data_sorted['Agent_Full_Name'], agent_data_sorted['Success_Rate'], 
             marker='o', color='red', label='Success Rate (%)')

# Annotations for success rates
for i, rate in enumerate(agent_data_sorted['Success_Rate']):
  line_plot.annotate(f"{rate}%", (i, rate), textcoords="offset points", xytext=(0, 5), ha='center', color='red')

# Labels and title
bar_plot.set_xlabel("Agent Full Name", fontsize=12)
bar_plot.set_ylabel("Total Successful Calls", fontsize=12)
line_plot.set_ylabel("Success Rate (%)", fontsize=12)
plt.title("Top-Performing Agents: Successful Calls and Success Rate (%)", fontsize=14)
bar_plot.tick_params(axis='x', rotation=45)

# Layout and show plot
plt.tight_layout()
plt.show()
      

4. Total Calls Made by Each Agent

This analysis shows the total number of calls made by each agent in descending order. By visualizing the call volume, we can evaluate the workload distribution across agents and identify high-call-volume agents who might require additional support or resources.

# Calculate total calls made by each agent, sorted in descending order
total_calls = df.groupby('Agent_Full_Name').size().reset_index(name='Total_Calls')
total_calls_sorted = total_calls.sort_values(by='Total_Calls', ascending=False)

# Display the result
print("Agent Full Name                Total Calls Made")
print("-------------------------------------------------")
for i, row in total_calls_sorted.iterrows():
  print(f"{row['Agent_Full_Name']:<30} {row['Total_Calls']:,.0f}")
      

5. Are Agent Ratings Correlated with Call Outcomes?

This analysis investigates whether there is a correlation between agent ratings and call outcome success rates. Understanding this relationship can help identify whether higher ratings are associated with better performance in handling successful calls.

# Step 1: Group by agent and calculate total calls, successful calls, and average ratings
agent_performance = df.groupby(['Agent_First_Name', 'Agent_Last_Name']).agg(
  total_calls=('Call_Outcome', 'size'),
  successful_calls=('Call_Outcome', lambda x: (x == 'Success').sum()),
  avg_rating=('Agent_Rating', 'mean')
).reset_index()

# Step 2: Calculate the success rate for each agent
agent_performance['success_rate'] = agent_performance['successful_calls'] / agent_performance['total_calls']

# Step 3: Calculate the correlation between agent ratings and success rate
correlation = agent_performance['avg_rating'].corr(agent_performance['success_rate'])

# Print the correlation value
print(f"Correlation between Agent Rating and Call Outcome Success Rate: {correlation:.4f}")
      

📊 Scatter Plot: Agent Rating vs. Call Outcome Success Rate

Below is a scatter plot visualizing the relationship between agent ratings and call outcome success rates. Each data point represents an individual agent, and the red regression line shows the trend.

          # Step 1: Add a new column for full names
          agent_performance['Agent_Full_Name'] = agent_performance['Agent_First_Name'] + ' ' + agent_performance['Agent_Last_Name']
          
          # Step 2: Create the scatter plot
          plt.figure(figsize=(12, 8))
          
          # Plot scatter and add a regression line
          sns.regplot(x='avg_rating', y='success_rate', data=agent_performance, logistic=False, scatter_kws={'s': 60, 'color': 'blue'}, line_kws={'color': 'red'})
          
          # Annotate each data point with the agent's full name in vertical orientation
          for i in range(len(agent_performance)):
              # Add a vertical space between points and the agent names by adjusting the y position
              y_offset = 0.02  # Adjust this value to control the amount of space
              plt.text(agent_performance['avg_rating'].iloc[i], 
                       agent_performance['success_rate'].iloc[i] + y_offset,  # Adding offset to y position
                       agent_performance['Agent_Full_Name'].iloc[i], 
                       fontsize=9, ha='center', va='bottom', rotation=90)
          
          # Add titles and labels
          plt.title('Relationship Between Agent Rating and Call Outcome Success Rate', fontsize=16)
          plt.xlabel('Average Agent Rating', fontsize=12)
          plt.ylabel('Call Outcome Success Rate', fontsize=12)
          
          # Show the plot
          plt.show()
        

From the scatter plot, it's evident that there is no strong linear relationship between agent ratings and success rates. While some high-rated agents are performing well, others do not seem to have a significantly higher success rate. This could imply that other factors might be influencing the success rate more than the agent rating.

Unlocking the Relationship Between Agent Ratings and Call Success Rates: A Data-Driven Perspective

In today’s fast-paced customer service landscape, businesses heavily rely on agent performance metrics to ensure optimal service quality. Two key indicators are often used to measure effectiveness: agent ratings (feedback provided by customers about their experience) and call success rates (the percentage of calls that achieve a desired outcome, such as issue resolution or sales conversion).

At first glance, one might assume that highly rated agents should naturally have a high success rate. However, our analysis of the relationship between these two metrics reveals a more complex and nuanced story—one that goes beyond just ratings and delves into factors like training, call handling techniques, and customer engagement strategies.

Statistical Breakdown: Understanding the Data

The scatter plot we analyzed presents a visualization of average agent ratings (x-axis) against call outcome success rates (y-axis). Several key takeaways emerge from this data:

1. Weak Positive Correlation

The red trend line in the graph indicates a slight upward slope, suggesting a weak positive correlation between agent ratings and call success rates. In statistical terms, this likely translates to a correlation coefficient of r ≈ 0.2 - 0.3, meaning that agent ratings do have an impact on success rates, but not significantly.

This insight challenges the conventional wisdom that simply having high agent ratings guarantees successful calls. Clearly, other factors influence whether a call is ultimately successful.

2. High Variance and Uncertainty

One of the most striking aspects of the visualization is the wide confidence interval (the shaded red region around the trend line). This indicates that there is considerable variability in call success rates, even among agents with similar ratings.

• Some agents with low ratings (<4.25) have very low success rates (~20%).
• However, others in the same rating range achieve success rates of 90% or more, proving that rating alone does not determine success.
• Even among highly rated agents (>4.27), there is significant variability, with some achieving high success and others struggling.

3. Outliers and Clusters

• Low-performing agents: Sophia Delacruz, Zoe Newman, and Elijah Hawkins have both low ratings and low success rates, indicating potential issues with their performance or customer interactions.
• High-performing but low-rated agents: Katrina Williams, Monique Lawrence, and Samuel Smith have exceptionally high success rates (above 80%) despite not having the highest ratings. This suggests that their strategies or experience might play a more critical role in success than customer feedback alone.
• Highly rated but underperforming agents: Some agents with high ratings (e.g., Drew Clay) still have moderate to low success rates, raising questions about whether customer satisfaction translates into tangible business outcomes.

Insights and Actionable Recommendations

Given these findings, how can businesses improve call success rates? Below are four key strategies that can enhance both agent performance and overall customer interactions:

A. Training & Coaching Enhancements

• Targeted Coaching for Low-Performing Agents: Identify agents with high ratings but low success rates and conduct call listening sessions to analyze where they struggle. Offer personalized coaching focused on objection handling, persuasive techniques, and emotional intelligence.
• Learn from High-Performing Agents: Study top performers (e.g., Katrina Williams, Monique Lawrence) to uncover patterns in their successful calls. Develop a best practices playbook based on their approaches, which can be shared across teams.
• Role-Playing and Simulation-Based Training: Conduct mock calls where agents practice handling difficult customer scenarios. Simulated training can be tailored based on real unsuccessful calls, helping agents learn from actual mistakes.

B. Process & Script Optimization

• Refining Call Scripts: Agents struggling with success rates might be using ineffective scripts. Implement A/B testing on different script variations to identify what works best. Allow for more flexibility in conversations, as rigid scripts may hinder natural customer engagement.
• Customer Segmentation & Adaptive Messaging: Different customer segments require different approaches. Train agents to adapt their tone, language, and strategy based on the customer’s profile and history.

C. Performance Metrics & Incentives

• Introduce Real-Time Performance Feedback: Implement dashboards where agents can see their success rates in real time, allowing them to self-adjust their performance. Provide instant coaching feedback from supervisors based on ongoing call analysis.
• Tie Incentives to Call Outcomes, Not Just Ratings: While agent ratings are important, companies should also reward agents based on actual success rates, ensuring they focus on driving real business outcomes. Consider performance-based bonuses that reward issue resolution and sales conversions.

D. Leveraging Technology & Automation

• AI-Powered Coaching: Use AI tools to analyze conversations in real-time, providing agents with live suggestions on how to improve their responses. Deploy speech analytics to detect customer frustration or hesitation, allowing for better in-call adjustments.
• Automated Call Summaries & Insights: Reduce agent workload by using AI-generated call summaries, allowing them to focus on higher-quality customer interactions rather than manual documentation. AI-driven insights can highlight common challenges faced by agents and suggest improvements over time.

Conclusion: The Bigger Picture

The relationship between agent ratings and call success rates is not as straightforward as it seems. While higher ratings may indicate better customer service, they do not necessarily translate into higher success rates. The data shows significant variability, with some agents defying expectations by performing exceptionally well despite moderate ratings—and others struggling despite strong customer feedback.

The key takeaway? A holistic approach is needed. By integrating advanced training, script flexibility, real-time feedback, performance-based incentives, and AI-driven tools, businesses can improve both customer experience and business success.

By shifting the focus beyond just ratings and diving deeper into actionable insights, organizations can empower agents to perform at their highest potential—ensuring both happy customers and successful outcomes. 🚀