Statapp-CANAL/Statapp-CANAL

This repo presents a Data Science project with Canal+ International on the personalization and performance of loyalty programs for prepaid subscribers in Africa. It showcases the structure, methods, and code behind the project, highlighting both technical implementation and business-oriented results. # Canal+ International - Data Science Project Personalization and Performance of Loyalty Programs in Africa --- ## Project Summary This project was carried out in collaboration with **Canal+ International** to analyze and improve the personalization and effectiveness of loyalty programs for prepaid subscribers in Africa. - **Context**: In Senegal, Canal+ operates with prepaid subscriptions where promotions play a crucial role in renewal behavior. The company needed to measure the true effectiveness of loyalty programs and move towards more personalized strategies. - **Data**: Over 10 million subscription records (2021-2023) from more than 500,000 subscribers. - **Approach**: - Built descriptive analyses to understand customer habits and promotion usage. - Designed a clustering framework (K-means) to segment subscribers into interpretable groups. - Quantified the effect of different promotions on each cluster using multiplicative renewal factors. - Validated the results on unseen data from November 2023. - **Results**: - Identification of distinct customer segments (loyal subscribers, promotion hunters, inactive users, etc.). - Clear evidence of which promotions truly drive re-subscriptions in which segments. - Business insights enabling Canal+ to reduce unnecessary promotions and target high-value subscribers more effectively. - **Impact**: - Personalized promotional strategies instead of blanket offers. - Better allocation of marketing resources. - Improved ROI and long-term customer loyalty. --- ## Code & Implementation This repository contains the code used to clean and process the data, perform clustering, and validate the results. The original datasets and the final insights of the study are **confidential** therefore not included here. ### Repository Structure ``` ├── data_operations/ │ ├── clustering/ │ │ └── Scripts for K-means clustering, optimal k selection, validation │ │ │ ├── tool_function/ │ │ └── Utility functions for data prepa …