Sales Analytics and Prediction Project

This project dives deep into customer sales data to uncover valuable insights for business decision-making. It leverages machine learning and time-series forecasting to predict customer churn,customer lifetime value (CLV) prediction, forecast product demand, and segment customers based on their purchasing behavior.

Table of Contents

• Project Overview
• Data
• Models and Analyses
• Power BI Dashboard
• Usage
• Limitations and Future Work

Project Overview

This repository contains Jupyter Notebooks and Python scripts for comprehensive sales analysis, addressing these key business challenges:

1. Customer Churn Prediction: Identifies customers at risk of leaving, enabling proactive retention strategies.
2. Demand Forecasting: Forecasts future product demand to optimize inventory levels and sales planning.
3. Customer Segmentation: Groups customers based on purchase behavior, enabling targeted marketing and personalization.
4. Customer Lifetime Value (CLV) Prediction: Estimates the long-term value of each customer to guide resource allocation and customer relationship management.

Data

• data_cleaned.csv: Cleaned dataset containing transaction-level information (customer ID, order ID, date, quantity, total price, product details).
• Data Source: You can download the data file from here: https://drive.google.com/file/d/10JEt8sq1NEhuv5AI7FDAmufsCy0qhVYC/view?usp=drive_link

Models and Analyses

Churn Prediction

• Model: XGBoost Classifier (optimized via GridSearchCV)
• Features:
  • Aggregated metrics: total orders, average order value, recency, frequency, unique products purchased
  • Calculated features: time since first/last purchase, change in purchase frequency, etc.
• Evaluation:
  • Metrics: Accuracy, precision, recall, F1-score, ROC-AUC, Precision-Recall Curve.
  • Interpretation: SHAP (SHapley Additive exPlanations) for understanding feature importance.

Demand Forecasting

• Model: Facebook Prophet
• Data: Aggregated daily (or other frequency) sales data.
• Optional: External regressors (holidays, promotions, etc.).

Customer Segmentation

• Model: K-Means Clustering
• Features: Standardized subset of relevant customer features.
• Evaluation: Elbow method, Silhouette analysis, interpretability of clusters.

Customer Lifetime Value (CLV)

• Model: BG/NBD (Beta Geometric/Negative Binomial Distribution) model from the lifetimes library
• Features: Recency, frequency, monetary value of transactions
• Output: Predicted CLV for each customer over a 12-month period.

Power BI Dashboard

This interactive dashboard provides a user-friendly way to explore and visualize the key insights from your sales data analysis. It's organized into three main sections, offering a holistic view of customer behavior, product performance, and overall sales trends.

Sales Report

1. Revenue Analysis:

   • Map: Geographic distribution of sales figures by city.
   • Animation: Evolution of customer segments based on total spending.
   • Line Chart: Total sales trend over time, broken down by month and year.
   • Table: Detailed sales data for each city, with summaries by family, range, and order counts.

2. Product Insights:

   • Cards: Top-selling products, sub-families, and ranges.
   • Table: Granular order data, including product counts and order values.
   • Pie Chart: Distribution of orders by day of the week.

Clients Analysis

1. Customer Overview:

   • Card: Number of customers
   • Card: City with the most customers
   • Histogram: Customer count by city
   • Interactive Map: Exploration of customer locations.
   • Timeline/Table: First and last order dates for each customer.

2. Churn Insights:

   • Pie Chart: Proportion of churned vs. not churned customers.
   • Table: Detailed customer data, including churn predictions.

Products Analysis

1. Product Overview:
   • Cards: Number of products and unique ranges.
   • Interactive Map: Sales by city and product.
   • Table: Product details with associated range, order counts, and total quantity sold.

Usage

1. Install Packages: pip install pandas numpy scikit-learn xgboost prophet lifetimes plotly ipywidgets
2. Jupyter Notebooks:
   • data_preparation.ipynb: Load and preprocess your data.
   • churn_prediction.ipynb: Train and evaluate the churn prediction model.
   • demand_forecasting.ipynb: Generate demand forecasts.
   • customer_segmentation.ipynb: Perform customer segmentation.
   • clv_prediction.ipynb: Calculate Customer Lifetime Value.
3. Examine Results: Look at visualizations, evaluation metrics, and SHAP explanations.
4. Apply Insights: Translate findings into actionable recommendations for your business.

Limitations and Future Work

• Explore Additional Models: Experiment with alternative algorithms for churn prediction and demand forecasting.
• Incorporate External Data: Enhance demand forecasts by including data on promotions, holidays, and other relevant factors.
• Expand Dashboard: Add more visualizations and interactivity to the Power BI dashboard.