Python project analyzing real usage behavior (calls and messages) of ConnectaTel customers across Mexico and Colombia, combining plan, user, and usage data to detect quality issues, outliers, and actionable customer segments.

Tools: Python · Pandas · NumPy · Seaborn · Matplotlib · Data Cleaning · Outlier Detection · Customer Segmentation

Project Context

Telecommunications companies capture large volumes of usage data, but without proper cleaning and segmentation it can be difficult to understand how customers actually use their services or which behaviors signal risk or opportunity.

This project analyzes ConnectaTel, a telecom provider operating in Mexico and Colombia, by combining plan details, customer profiles, and real call-and-message usage records through 2024. The analysis focuses on identifying data quality issues, building per-customer usage statistics, detecting outliers, and segmenting customers by age and consumption level.

Business Objective

The objective was to integrate and clean three related datasets to build a clear, reliable, and actionable view of customer usage behavior in order to:

• Identify and resolve data quality issues such as sentinel values, invalid dates, and missing records.
• Build a per-customer statistical summary of calling and messaging behavior during 2024.
• Detect outliers that could indicate unusual usage or data entry errors.
• Segment customers by age group and usage level to reveal differentiated behavior patterns.
• Translate the findings into recommendations that support commercial offers and customer experience decisions.

<aside> 💡

Key analytical question: Which customer segments show the highest and lowest usage of calls and messages, and what patterns or outliers should inform ConnectaTel’s commercial strategy?

</aside>

Dataset & Methodology

Dataset

The analysis combined three related data sources covering customer plans, profiles, and real usage during 2024. The main fields used included:

• Plan details: price, included minutes and GB, and extra-usage costs
• Customer profile: age, city, registration date, contracted plan, and churn status