Predictive Operations Control Center on Shiny Dashboards

This project combines a Demand Forecasting and Customer Analytics Dashboard with a Machine Learning-Driven Product Price Prediction Tool, creating an integrated suite for strategic sales, inventory, and pricing optimization for a bicycle retail business. The solution operationalizes advanced analytics into user-friendly Shiny applications, delivering immediate, data-driven insights to both management and sales teams.

Demand Forecasting and Customer Analytics Dashboard

This project demonstrates a high-impact, interactive Shiny application designed for rapid sales analysis and predictive demand forecasting within a simulated bicycle retail business. The solution operationalizes advanced analytics into a user-friendly dashboard, providing immediate, actionable insights.

Originally developed as coursework at Business Science University, the application extends core business analysis with robust data science methodologies and a professional-grade user experience (UX).

Business Value & Impact (For Stakeholders)

The dashboard enables stakeholders to quickly assess market performance, identify sales opportunities, and proactively manage future demand:

• Strategic Focus: Rapidly evaluate revenue concentration by geographic market to concentrate sales efforts and optimize resource allocation.
• Performance Monitoring: Utilize intuitive health and wealth metrics (e.g., sales above/below $10k, volume thresholds) to flag areas requiring immediate attention or celebrating success.
• Proactive Planning: Leverage integrated forecasting capabilities to visualize future trends and hedge against potential opportunities or risks in inventory management and sales targets.
• Actionable Insights: Drill down into specific customer segments to understand purchasing behavior and identify upsell potential (e.g., Mountain/Road bike ratio analysis).

Technical Overview & Methodology (For Data Scientists)

The application leverages a modern R-based tech stack to deliver speed, versatility, and aesthetic appeal:

• Frontend: Developed using the shiny package for reactive web interfaces, enhanced with custom CSS for professional aesthetics (including light/dark themes and dynamic colored indicators).
• Backend & Data Layer: Data is sourced from an SQLite database file. The application manages data parsing and interactivity efficiently.
• Forecasting Methodology: Employs machine learning (ML) models for time-series forecasting, specifically chosen over traditional methods (like ARIMA) to showcase faster performance and greater versatility across various time divisions (daily, weekly, monthly).
• Feature Engineering: The timetk package was utilized to derive extensive time series features for the ML models, significantly boosting real-time prediction performance.
• Visualizations: Includes loess smoothers within time-series plots for localized trend detection and thematic state mapping to visualize revenue density.

Key Application Features

• Interactive Forecasting: Ability to predict and visualize current and future trends at flexible time intervals.
• Geographic Analysis: A choropleth map visualization of US states shaded by revenue size, highlighting top-performing markets.
• Dynamic Metric Indicators: Visual health metrics for sales volume (Danger/Warning/Healthy zones) and revenue metrics (e.g., under $5K, under $10K, $10K+).
• Customer Segmentation: Mountain-to-Road bike ratio labels used to signal warning zones/upsell potential for specific customer segments.

Machine Learning-Driven Product Price Prediction Tool

This project showcases a practical application of machine learning developed to empower sales teams with immediate, accurate pricing capabilities. By providing a self-service tool, it minimizes friction in the sales cycle and streamlines product management workflows.

Business Value & Impact (For Stakeholders)

• Empowered Sales Teams: Provides sales representatives and managers with an autonomous, field-ready tool to generate accurate price estimates for standard products instantly, removing reliance on central pricing teams.
• Operational Efficiency: Minimizes the need for product management intervention on routine pricing inquiries, freeing specialists to focus on strategic initiatives.
• Strategic Pricing: Ensures consistent and data-informed pricing for new product configurations, helping the organization strategically fill product line gaps and maximize revenue potential.
• Accelerated Sales Cycle: Quick, reliable pricing estimates facilitate faster deal closure and improved customer response times.

Technical Overview & Methodology (For Data Scientists)

The application leverages a predictive machine learning model to provide real-time pricing guidance based on existing product data:

• Core Functionality: The tool allows users to input parameters for a hypothetical new bike model. It utilizes a trained machine learning model to predict an optimal price point based on the existing product portfolio data.
• Interactive Interface: Built with an interactive Plotly graph that visualizes the existing product landscape. The specific predicted price point for the user’s input is dynamically highlighted in green, providing immediate visual context.
• User Input Handling: The application handles user inputs efficiently, requiring standardized formatting (e.g., camel case for model names) to interface correctly with the underlying data model.
• Model Integration: The focus is on operationalizing a machine learning model into a reliable, front-end application that delivers analytical value directly to end-users.

Key Application Features

• Dynamic Prediction: Predicts prices of new product configurations in real-time based on user-defined feature parameters.
• Interactive Visualization: Uses Plotly for rich, interactive graphing that contextualizes the predicted price within the current market landscape.
• Self-Service Enablement: A user-friendly interface designed specifically for non-technical field personnel.