⚡ Dynamic Pricing Optimization: Switchback Experiment Live App Link
Figure 1: Experiment Control and KPI Overview
Figure 2: Time-Series Analysis and Statistical Significance
📌 Project Overview
This project is a Mock Experimentation Platform designed to simulate a ride-hailing marketplace (like Bolt). It demonstrates how to solve Network Interference problems in two-sided marketplaces using Switchback Testing.
The Problem: Why not standard A/B Testing?
In a ride-hailing app, riders and drivers share a finite pool of liquidity.
If we assign User A to a "Low Price" group and User B to a "High Price" group, User A might book the last available driver.
User B now has no driver available, not because of the high price, but because User A "stole" the supply.
Result: The Control group is contaminated by the Treatment group (Spillover Effect / SUTVA Violation).
The Solution: Switchback Testing
Instead of randomizing users, we randomize time.
Method: The platform alternates between Control (Standard Pricing) and Treatment (Dynamic Surge) algorithms in 30-minute windows.
Analysis: Metrics are aggregated by time window, and a robust T-test is performed on the window means to determine statistical significance.
🛠️ Tech Stack
Language: Python 3.9+
Backend: FastAPI (Simulation Engine & Statistical API)
Frontend: Streamlit (Interactive Dashboard)
Analysis: SciPy (Welch’s T-test), Pandas, NumPy
🚀 How to Run Locally
- Setup Environment
Open your terminal and create a virtual environment:
python -m venv venv .\venv\Scripts\activate
python3 -m venv venv source venv/bin/activate
- Install Dependencies
pip install -r requirements.txt
- Run the Backend (Simulation Engine)
Open a new terminal, activate venv, and run:
uvicorn backend.main:app --reload
The API will be available at https://www.google.com/search?q=http://127.0.0.1:8000
- Run the Frontend (Dashboard)
Open a second terminal, activate venv, and run:
streamlit run frontend/app.py
📊 Key Metrics Defined
OCR (Order Completion Rate): Completed Rides / Total Requests. This is the primary metric for liquidity.
GMV (Gross Merchandise Value): Total revenue generated. Used as a guardrail metric.