Customer Satisfaction Analysis

Solution

Here are the data types for each survey response:

• (a) "How satisfied are you with our service?" (Options: Very Dissatisfied, Dissatisfied, Neutral, Satisfied, Very Satisfied)
  • Data Type: Ordinal. The options have a meaningful order or rank (Very Satisfied is better than Satisfied, which is better than Neutral, etc.). However, the differences between the categories are not necessarily equal or quantifiable (e.g., the jump in satisfaction from 'Dissatisfied' to 'Neutral' isn't necessarily the same as from 'Satisfied' to 'Very Satisfied').
• (b) "How many times have you contacted our customer support centers in Hyderabad, Warangal, or Vijayawada in the last month?" (Numeric input)
  • Data Type: Discrete (Numerical). This is a count of occurrences. The values are whole numbers (0, 1, 2, 3, etc.), and you can't have a fraction of a contact. It's quantitative and countable.
• (c) "What is your primary reason for contacting support?" (Open-text field, later categorized into themes like 'Billing Issue', 'Technical Problem during IPL matches', 'Router Configuration', 'Connection Drops during Rains')
  • Data Type (after categorization): Nominal (Categorical). The categorized themes ('Billing Issue', 'Technical Problem during IPL matches', etc.) are distinct categories with no inherent order or ranking among them. One reason is not inherently "higher" or "lower" than another; they are just different types of issues. The raw open-text data itself is unstructured text.

The neighborhoods mentioned (Gachibowli, Miyapur, Begumpet, Kukatpally, Uppal) or customer support centers (Hyderabad, Warangal, Vijayawada) are contextual information or potential segments for later analysis, but they don't change the fundamental data type of the responses themselves.

Solution

Here's how I would present findings for each data type to Nagarjuna FiberNet's executives at Cyber Towers:

(a) Satisfaction Level (Ordinal Data)

• Descriptive Statistics:
  • Frequency Distribution & Percentages: For each category (Very Dissatisfied to Very Satisfied).
  • Mode: The most frequent satisfaction level.
  • Median: The middle satisfaction level when responses are ordered. This is a key measure of central tendency for ordinal data.
  • (Optionally, if assigning numerical scores like 1-5: Mean satisfaction score, but with the caveat that intervals aren't truly equal).
• Visualizations:
  • Bar Chart: Showing the frequency or percentage of each satisfaction level. This is very intuitive.
  • Stacked Bar Chart: Useful for comparing satisfaction distributions across different segments (e.g., service tiers, but see limitations).
  • Diverging Stacked Bar Chart: Good for visualizing responses around a neutral point.
• Limitations:
  • Calculating a mean is controversial for ordinal data as it assumes equal intervals. If used, this must be clearly stated.
  • When comparing service tiers (e.g., "Dasara Special Plan" vs. "Raja Plan"), if one tier has a very small number of respondents, its satisfaction distribution might not be statistically reliable or representative. Small sample sizes can lead to misleading comparisons.
  • The "Dasara Special Plan" might attract a specific type of customer (e.g., more price-sensitive, short-term) whose baseline satisfaction expectations differ from premium "Raja Plan" users, making direct comparison tricky without further context.

(b) Number of Support Contacts (Discrete Data)

• Descriptive Statistics:
  • Frequency Distribution: Number of customers for each contact count (0, 1, 2, etc.).
  • Mean, Median, Mode: To describe the typical number of contacts. The mean can be skewed by a few customers with very high contact counts.
  • Range, Interquartile Range (IQR), Standard Deviation: To describe the spread or variability in contact numbers.
• Visualizations:
  • Histogram or Bar Chart (for discrete counts): Showing the distribution of contact frequencies.
  • Box Plot: To show the distribution, median, quartiles, and potential outliers, especially when comparing across segments.
• Limitations:
  • The mean can be misleading if the data is highly skewed (e.g., most customers have 0-1 contacts, but a few have 10+). Median might be better.
  • Outliers (customers with exceptionally high contact numbers) should be investigated – they might represent severe unresolved issues or specific user types.

(c) Primary Reason for Contact (Nominal Data - after categorization)

• Descriptive Statistics:
  • Frequency Counts & Percentages: For each categorized reason (e.g., 'Billing Issue', 'Technical Problem during IPL matches').
  • Mode: The most common reason for contacting support.
• Visualizations:
  • Bar Chart: Displaying the frequency or percentage of each reason. Easy to compare categories.
  • Pareto Chart: A bar chart with reasons sorted by frequency, also showing a cumulative frequency line, to highlight the most significant reasons (80/20 rule).
  • Pie Chart: Can be used if there are few categories (e.g., 3-5), but generally less effective than bar charts for comparison, especially with many categories like 'Technical Problem during IPL matches' vs 'Connection Drops during Rains'.
• Limitations:
  • The quality of insights depends heavily on the accuracy and consistency of the categorization of open-text responses. Ambiguous or overlapping categories can be problematic.
  • The "Other" category, if large, can hide important specific issues.
  • When comparing service tiers, certain issues might be more prevalent in one tier due to its features or customer base (e.g., more 'Router Configuration' issues with a plan that allows custom routers).

Solution

To analyze relationships between satisfaction levels and the number of support contacts, especially segmented by districts (Rangareddy, Medchal, Krishna), and to help Mr. Suresh Reddy develop targeted strategies, I would undertake the following:

1. Data Preparation and Segmentation:
   • Ensure data is clean and customer districts are accurately mapped.
   • Segment the data by district: Rangareddy, Medchal, Krishna, and potentially other relevant geographical units.
2. Relationship Analysis within Each District:
   • Cross-Tabulation (Contingency Tables): Create tables showing the frequency of customers for each combination of satisfaction level (ordinal) and number of support contacts (discrete, possibly binned into groups like 0, 1, 2, 3+ for clarity). This should be done for each district separately and overall. Example: For Rangareddy district, how many "Very Dissatisfied" customers had 3+ contacts?
   • Visualizations:
     • Grouped or Stacked Bar Charts: Plot satisfaction levels, with bars segmented by the number of support contacts (or vice-versa). This helps visually compare distributions. For instance, a chart showing the distribution of support contact counts for "Very Satisfied" customers versus "Very Dissatisfied" customers.
     • Box Plots: Show the distribution of support contacts for each satisfaction level. This can reveal if, for example, lower satisfaction levels are associated with a higher median or wider spread of contact numbers.
   • Correlation (with caution): If satisfaction levels are numerically encoded (e.g., 1-5), calculate a correlation coefficient (like Spearman's rank correlation, suitable for ordinal data) between satisfaction score and number of contacts. This gives a single metric of association, but nuance from cross-tabs is important.
   • Statistical Tests: A Chi-squared test for independence can be used on the cross-tabulation to see if there's a statistically significant association between satisfaction level and categories of contact frequency.
3. Identify Patterns and Key Drivers of Dissatisfaction:
   • Look for trends: Do customers with more support contacts consistently report lower satisfaction across all districts, or is this pattern stronger in specific districts like Medchal compared to Krishna?
   • Combine with "Reason for Contact" data: For dissatisfied customers with high contact counts in a specific district, what are their primary support reasons? For example, are 'Connection Drops during Rains' in Uppal (Medchal district) strongly linked to high contacts and low satisfaction?
4. Develop Targeted Service Improvement Strategies for Mr. Suresh Reddy:
   • District-Specific Action Plans:
     • If Rangareddy district shows high dissatisfaction linked to 'Technical Problems during IPL matches' and frequent contacts, strategies could include proactive network capacity upgrades in Rangareddy before major sporting events, or dedicated support teams for IPL-related issues.
     • If Medchal district has high dissatisfaction due to 'Connection Drops during Rains' and multiple contacts, focus on weather-proofing infrastructure in vulnerable areas of Medchal before the next monsoon/festive season.
   • Proactive Support for High-Contact Customers: Identify customers with multiple contacts who are still 'Neutral' or 'Dissatisfied'. Implement proactive outreach or case escalation for them.
   • First-Contact Resolution (FCR) Focus: If data shows that multiple contacts (even for the same issue) correlate with lower satisfaction, emphasize improving FCR rates at support centers in Hyderabad, Warangal, and Vijayawada. This might involve better training or empowering support staff.
   • Channel Optimization: If the Telugu mobile app is a primary source of feedback, ensure its support features are robust and issues reported via app are resolved efficiently.
   • Pre-Festive Season Preparedness: Based on common issues in high-complaint districts (like Kukatpally), implement preventative maintenance or temporary capacity boosts before the festive season to mitigate known problems.
5. Presentation to Mr. Reddy:
   • Summarize findings with clear visuals focusing on key relationships in problematic districts.
   • Highlight the most impactful areas for improvement.
   • Provide data-driven recommendations for resource allocation for service improvement.