📦 Chainpulse : Supply Chain Bottleneck Analytics

🚀 Business Use Case : Predictive Supply Chain Risk & Bottleneck Analytics

The primary goal of this project is to provide logistics managers and procurement officers with a predictive engine to identify "at-risk" shipments and inventory bottlenecks before they disrupt operations. This project focuses on identifying and analyzing supply chain bottlenecks using data-driven insights. By comparing actual shipping times against scheduled timelines, this analysis pinpointed delays across different shipping modes, markets, and product categories. The goal is to provide actionable intelligence to optimize logistics efficiency and enhance customer satisfaction.

Wider Industry Applications: The logic used in this project can be adapted for several high-value supply chain environments:

• Manufacturing (JIT): Predicting delays in raw material arrivals to prevent production line downtime.
• Cold Chain Logistics: Identifying high-risk routes for temperature-sensitive pharmaceuticals or food products.
• Retail Inventory Management: Analyzing "Order Gaps" and "Stock-out Risks" to optimize safety stock levels across global warehouses.
• Last-Mile Delivery: Predicting delivery failures or delays in urban zones based on courier density and historical traffic patterns.
• Freight Forwarding: Using ensemble models to predict transit time variability across sea, air, and land freight.

---

📊 Dataset Description

The analysis uses the DataCo Smart Supply Chain Dataset, which includes typical supply chain activities such as provisioning, production, sales, and commercial distribution.

• Key Features Used:
  • Days for shipping (real): The actual time taken to deliver.
  • Days for shipment (scheduled): The promised delivery time.
  • Shipping Mode: Standard Class, First Class, Second Class, Same Day.
  • Market: Pacific Asia, USCA, Africa, Europe, LATAM.
  • Order Status: Status of the order delivery.
  • Product Price & Category Name: Details regarding the items sold.

---

🛠️ The Approach & Methodology

The problem was approached as a Risk Estimation and Bottleneck Classification challenge. The workflow followed these key phases:

• Data Acquisition & Cleaning: Large-scale supply chain datasets were processed to handle missing logistics coordinates and normalize temporal fields like Days for shipping (real).
• Feature Engineering: Calculated critical KPIs such as Shipment Delay (Actual vs. Scheduled) and Internal Processing Time to isolate operational friction.
• Bottleneck Identification: Using machine learning drivers, I identified that Shipping Mode, Order Region, and Product Category are the highest predictors of delays.
• Model Architecture: The system utilizes a modular approach, offering a Market Dashboard for macro logistics trends, an Individual Shipment Risk Score for real-time tracking, and Batch Processing for warehouse-wide audit analysis.
• Deployment: The engine is hosted via a high-fidelity Streamlit web application, featuring an intuitive, dark-themed interface for supply chain command centers.

---

⚙️ Data Pipeline

1. Data Cleaning: * Handling missing values in Customer Zipcode and Product Description.
   • Converting order_date and shipment_date to datetime objects.
   • Stripping whitespace from column headers for consistent indexing.
2. Feature Engineering:
   • Shipment_Delay: Calculated as Real Days - Scheduled Days.
   • Is_Bottleneck: Boolean flag indicating if a delay occurred.
   • Internal_Processing_Time: Time difference between order date and shipment date.
3. Exploratory Data Analysis (EDA):
   • Distribution of shipping days across different markets.
   • Correlation analysis between Product Price and Shipping Speed.
   • Proportion analysis using Donut charts and Countplots.

---

🚀 Key Findings from Analysis

Based on the notebook execution, the following findings were discovered:

1. High Bottleneck Rate: Approximately 57.28% of all orders experienced a bottleneck (Actual days > Scheduled days).
2. Mode Inefficiency: "Second Class" shipping showed the highest average shipment delay (~1.99 days), while "Standard Class" was the most reliable relative to its schedule.
3. Risk Categories: Products in the Golf Bags & Carts and Soccer categories are most prone to shipping delays.
4. Market Variance: Specific markets like LATAM and Pacific Asia show wider variance in shipping times compared to USCA.

💡 Key Insight: Feature Importance analysis reveals that the combination of Shipping Mode and Product Category accounts for ~40% of the predictive weight, suggesting that logistical routing is the primary driver of supply chain friction.

---

📈 Visualizations Included

• Bottleneck Proportion: A donut chart showing the split between on-time and delayed orders.
• Shipping Mode Impact: Countplots comparing bottlenecks across various logistics tiers.
• Market Distribution: Violin plots showing the density and quartiles of shipping durations per global market.
• Regression Analysis: Scatter plots identifying if higher product prices correlate with faster shipping.

🔍 Deep Dive: Operational Indicators

• The "Second Class" Paradox: Analysis reveals that "Second Class" shipping often experiences higher variance than "Standard Class," suggesting a need for route optimization in mid-tier logistics.
• Regional Variance: The Violin Plots highlight that the LATAM market has a wider "tail" in shipping days, indicating a higher frequency of extreme delays compared to the USCA market.
• Pricing Correlation: Regression analysis indicates a low correlation between product price and shipping speed, debunking the myth that higher-value items are inherently expedited by default system logic.

---

📂 Project Structure

├── scba.ipynb                       # Main Jupyter Notebook with analysis & plots
├── Data/
│   └── DataCoSupplyChainDataset.csv  # Raw Input Data (Source)
├── Processed_Supply_Chain_Data.csv   # Cleaned data exported after analysis
└── app.py                            # Streamlit Multi-page Web Application
└── requirements.txt                  # Project Dependencies
└── README.md                         # Project documentation

---

🛠️ Tech Stack

• Language: Python
• Libraries: * Pandas: Data manipulation and cleaning.
  • Matplotlib & Seaborn: Advanced data visualization.
  • NumPy: Numerical operations.
  • Streamlit (Imported for future dashboard integration).

---

🏗️ How to Run

1. Clone the Repository:

   git clone https://github.com/MTank76/Supply-Chain-Bottleneck-Analysis.git
   cd Supply-Chain-Bottleneck-Analysis

2. Install Dependencies

   pip install -r requirements.txt

3. Run the Notebook: Open scba.ipynb in Jupyter Notebook or VS Code and run all cells.

4. Fire up the Dashboard**

   streamlit run app.py

Quick Start: pip install -r requirements.txt && streamlit run app.py

🤝 References

• Dataset Source: UDataCo SMART SUPPLY CHAIN FOR BIG DATA ANALYSIS / Kaggle
• Inspired by: Stephan M. Wagner - Bottleneck identification in supply chain networks