Wingman

Inspiration

Travel disruptions, especially flight cancellations and delays, are a common inconvenience, often leaving travelers confused about their refund eligibility. Our team drew inspiration from personal experiences—moments spent waiting at airline counters or sifting through complicated refund policies. We asked ourselves: "Why can’t travelers instantly know their chances of receiving a refund?" From that question, Wingman’s Refund Probability Estimator was born. Our goal was to give travelers a quick, data-driven answer to a crucial question during stressful travel moments.

What We Learned

Throughout the development of Wingman’s Refund Probability Estimator, we gained valuable insights into both technical challenges and user needs. Key lessons included:

• Simplicity is key: Travelers dealing with disruptions need fast, clear answers without complexity.
• Data quality drives accuracy: Reliable datasets are essential for building meaningful predictive models.
• Modeling requires careful consideration of features: Variables like delay duration and flight type (domestic vs. international) significantly impact refund likelihood.
• User empathy matters: Presenting refund probabilities in an intuitive way (e.g., through visual progress bars) improves user trust and understanding.

How We Built It

We began by defining core user needs: a straightforward way to estimate refund chances based on flight details. Our process included:

• Data Generation and Modeling: Lacking accessible real-world refund data, we created a comprehensive mock dataset to train a machine learning model using scikit-learn. We prioritized realism by considering factors like delay duration and international vs. domestic flights.
• Backend Development: We built a Flask API to serve the model’s predictions and a Node.js/Express server to handle requests from the front end.
• Frontend Development: Using React (with Vite), we designed a clean interface where users can input flight delay details and receive refund probability estimates. We included a progress bar for visual clarity and ensured inputs only accepted valid values (e.g., whole numbers for minutes).
• UI/UX Improvements: We focused on responsive design, added logo branding and chose modern fonts to enhance the overall user experience.

Challenges We Faced

Like any project, Wingman wasn’t without hurdles. Some key challenges included:

• Data scarcity: Refund data isn’t publicly available, so we had to rely on synthetic data generation and careful feature engineering.
• Model tuning under time constraints: Balancing model accuracy with development speed required thoughtful decisions.
• Integration complexity: Ensuring smooth communication between the front, Node.js server, and Flask API was challenging, especially when handling cross-origin requests and model reloading.
• Time pressure: With limited time, we focused on delivering a polished core feature (refund estimation) while setting aside additional features for future iterations.

Why Wingman Stands Out

Wingman’s Refund Probability Estimator provides immediate, data-driven insights that travelers can rely on during disruptions. Unlike tools that only show policy information, Wingman delivers a personalized refund probability, helping users make informed decisions faster. Its intuitive interface, quick predictions, and clear visuals set it apart as a practical solution for travelers in distress.

Final Thoughts

This project was a journey of problem-solving, learning, and teamwork. We’re proud of what we built—Wingman empowers travelers to face disruptions with more information and less stress. While there’s room to expand (adding live flight data or broader features), our focus on refund probability provides real value to users facing uncertain travel situations.

Built With

• css
• figma
• flask
• github
• html
• javascript
• node.js
• pandas
• python
• react
• scikit-learn
• vscode