Revenue Optimization System: Leveraging Association Rule Mining and Logistics Simulation to boost Average Order Value (AOV) and reduce operational costs.
This project is an end-to-end data product designed to bridge the gap between Data Science and Business Strategy. By analyzing over 100,000 orders from the Olist E-Commerce dataset, the system identifies high-value cross-selling opportunities (product bundles) and simulates the logistics cost savings of shipping these items together.
Unlike traditional analysis, this engine provides actionable financial metrics:
- 10.1x Lift Score identified between specific categories (e.g., Watches & Audio).
- 24% Reduction in shipping costs via package consolidation.
- Direct Margin Impact calculation for every proposed bundle.
In the competitive e-commerce landscape, two major challenges erode profitability:
- Low Average Order Value (AOV): Customers tend to purchase single items, leading to high customer acquisition costs relative to revenue.
- High Logistics Costs (OPEX): Shipping items individually results in inefficient use of packaging and logistics networks, increasing "Last Mile" delivery costs.
The Challenge: How can we scientifically determine which products should be bundled together to simultaneously drive sales volume and optimize operational efficiency?
This project utilizes a hybrid methodology combining Data Analytics, Industrial Engineering, and Business Development:
- Technique: Used FP-Growth Algorithm (Association Rule Mining) on a transactional dataset.
- Process: Filtered out single-item "noise" orders to focus on multi-item baskets, creating a sparse matrix to identify hidden purchasing patterns.
-
Result: Identified strong relationships (e.g., Baby Products
$\rightarrow$ Cool Stuff/Gadgets with a Lift of 4.99).
- Engineering Logic: Simulated the physical consolidation of products.
- Calculation: Compared the cost of separate shipments vs. a single consolidated package based on weight and volume metrics.
- Result: Calculated a net saving of 7.84 BRL per order for top-performing bundles.
- Business Logic: Converted technical metrics into financial KPIs.
- Outcome: Proposed a "Weekend Bundle Strategy" where the logistics savings are reinvested into customer discounts or profit margins.
The project is built with a modular Python architecture:
- Language: Python 3.10+
- Data Manipulation:
Pandas,NumPy - Machine Learning:
MLxtend(Apriori & FP-Growth algorithms) - Visualization:
Matplotlib,Seaborn(for exploratory analysis) - Architecture:
data_loader.py: Automated data ingestion from remote repositories.processor.py: Data cleaning, sparsity reduction, and filtering.market_basket.py: Rule mining engine.logistics.py: Cost simulation and margin analysis engine.
Follow these steps to set up the project on your local machine:
1. Clone the Repository
git clone https://github.com/Farslan-x/Strategic-Marketplace-Growth-Engine.git
cd Strategic-Marketplace-Growth-Engine- Dynamic Pricing Engine: Developing an algorithm to reinvest a portion of the logistics savings into customer-facing "Dynamic Discounts" to increase conversion rates.
- Geospatial Analysis: Tailoring bundle recommendations based on regional data (e.g., prioritizing Beachwear bundles in Rio de Janeiro vs. Office Supplies in Sao Paulo).
- API Deployment: Encapsulating the model within a FastAPI microservice for real-time integration with live e-commerce platforms.