In this project we try to Optimize Revenue Management for Sustainable Water Distribution: QUATECH Senegal's Customer Data Analysis to Address the Decline in Water Development Fund Revenues
# Optimizing Revenue Management for Sustainable Water Distribution
### QUATECH Senegal – Customer Data Analysis
## Overview
This project analyzes the decline in **Water Development Fund (FDE)** revenues at QUATECH Senegal despite increasing water production.
A **data-driven approach** is used to uncover key drivers of revenue decline and support strategic decision-making.
---
## Problem Statement
Why are FDE revenues decreasing despite increased water production and high payment compliance?
---
## Objectives
- Segment subscribers based on **consumption and payment behavior**
- Analyze **billing and payment trends**
- Identify key factors driving **FDE revenue decline**
- Build predictive models for **revenue estimation**
---
## 📊 Dataset
- **Size:** 1.14M records, 23 features
- **Period:** 2014 – 2019
- **Region:** Saint-Louis (Region 4)
- Includes:
- Consumption (cubage)
- Billing & payments
- Customer category (Private/Admin)
- Financial metrics (FDE, VAT, FNE)
---
## Data Preprocessing
- Renamed variables for clarity
- Converted date and monetary fields
- Handled missing values (`PAY_DATE` retained for delay analysis)
- Removed extreme anomalies (outliers in FDE)
- Encoded categorical variables
- Feature engineering:
- `PAYMENT_DELAY`
- Time-based features
---
## Key Insights
- FDE revenue **peaked in 2017**, declined sharply after
- **94.5% pay on time**, but revenue still drops
- **Administrative customers consume ~87.5%** of water
- Revenue concentrated in few regions:
- Bakel (**52.6%**)
- Foundiougne (**13.5%**)
- Birkelane (**12.1%**)
- Average payment delay ≈ **30 days**
---
## Methodology
### 1. Exploratory Data Analysis (EDA)
- Trend analysis (FDE over time)
- Payment behavior analysis
- Regional contribution analysis
### 2. Dimensionality Reduction
- **PCA** to identify key behavioral patterns:
- Axis 1 → Consumption & billing
- Axis 2 → Payment behavior
### 3. Modeling
- Regression Models:
- **Ridge** (best performance)
- Lasso
- …