Strategix

• 

Inspiration

Strategix was inspired by a common challenge in banking campaigns: teams often contact thousands of customers without knowing who is actually likely to respond. This creates wasted calls, higher campaign costs, and missed opportunities. We wanted to build a tool that helps banks move from broad outreach to smarter, data-driven decision-making.

The idea behind Strategix is simple: call smarter, not harder. By combining machine learning with an interactive campaign dashboard, we wanted to help managers understand customer behavior, prioritize the best leads, and estimate the business impact of different campaign strategies.

What it does

Strategix is an intelligent campaign decision platform for bank term-deposit marketing. It helps campaign teams identify which customers are most likely to subscribe and turns model predictions into practical business actions.

Strategix can:

• Score customers by predicted subscription probability
• Categorize leads as Hot Lead, Warm Lead, or Low Priority
• Filter and export ranked customer call lists
• Simulate campaign ROI using call cost, revenue, and probability thresholds
• Show estimated conversions, revenue, costs, calls saved, and profit
• Explain important model drivers through feature importance
• Test custom customer profiles with live prediction results
• Generate a manager-ready campaign recommendation brief

How we built it

We built Strategix as a full-stack application with a React frontend and a FastAPI backend.

The backend is written in Python using FastAPI. It loads the bank marketing dataset, prepares the features, trains a Random Forest Classifier, and exposes API endpoints for the React app. The machine-learning pipeline uses pandas, NumPy, and scikit-learn.

The frontend is built with React and Vite. We designed a modern dashboard interface with interactive pages for campaign analysis, lead prioritization, ROI simulation, model insights, profile testing, and manager recommendations. The UI uses custom CSS, responsive layouts, hover interactions, and visual summaries to make the experience feel polished and easy to understand.

One important modeling decision was removing the duration feature from the training data. Since call duration is only known after a call happens, including it would create data leakage. Strategix focuses on pre-call decision-making, so the model only uses information available before outreach.

Core technologies:

Frontend: React, Vite, CSS
Backend: FastAPI, Uvicorn
Machine Learning: pandas, NumPy, scikit-learn
Model: Random Forest Classifier
Dataset: Bank marketing campaign dataset

Challenges we ran into

One challenge was turning machine-learning predictions into something useful for a real campaign manager. A probability score alone is not enough, so we added lead priority categories, ROI simulation, threshold controls, and a manager brief to make the model output actionable.

Another challenge was avoiding data leakage. The original dataset includes duration, which is highly predictive but not available before making a call. We removed it so the model would reflect a realistic pre-call campaign scenario.

We also worked through frontend design challenges. We wanted the app to feel modern and engaging, but still professional for a banking use case. This meant balancing color, charts, interactivity, and clean layout without making the interface feel overwhelming.

Finally, connecting the React frontend with the FastAPI backend required careful API design. The frontend needed enough data to power charts, filters, ROI calculations, and live predictions while keeping the experience smooth for users.

Accomplishments that we're proud of

We are proud that Strategix is more than just a model demo. It is a complete decision-support tool that connects machine learning to business value.

Some accomplishments we are especially proud of:

• Built a full-stack React and FastAPI application
• Created a realistic pre-call prediction pipeline
• Designed an interactive lead prioritization workflow
• Added ROI simulation to show business impact
• Built a live customer profile prediction tool
• Created a manager-ready recommendation page
• Improved the UI with polished styling, hover details, and responsive layouts
• Made the app easy to demo and understand for both technical and non-technical audiences

What we learned

We learned how important it is to connect data science with real user decisions. A model can produce accurate predictions, but the value comes from making those predictions understandable and actionable.

We also learned more about preventing data leakage, designing machine-learning workflows, and explaining model results in a business-friendly way. On the frontend side, we learned how to create a more polished React interface with interactive charts, filters, exports, and live API calls.

Most importantly, we learned that a strong data product needs both good analytics and good user experience. Strategix helped us bring those two sides together.

What's next for Strategix

Next, we would like to expand Strategix into a more complete campaign management platform.

Future improvements could include:

• Cloud deployment
• User authentication for campaign teams
• CRM integration
• Real-time campaign tracking
• A/B testing for outreach strategies
• More advanced explainability for individual predictions
• Campaign history and performance comparison
• Automated recommendation tuning based on early campaign results

Our long-term vision is for Strategix to become a practical tool that helps banks plan smarter campaigns, reduce wasted outreach, and improve customer targeting.

Built With

• css
• css-backend:-fastapi
• fastapi
• numpy
• pandas
• react
• scikit-learn
• uvicorn
• uvicorn-data/ml:-pandas
• vite