Foot Traffic Finder

Search for locations and dates to get a prediction of local foot traffic
A result will appear under calendar giving concise forecasts for local foot traffic
Legend at bottom left of screen that explains the point values that show up on the map. Shows the range for busy, moderate, and busy traffic
The map shows points with varying sizes based on amount of predicted foot traffic

Inspiration

San Francisco’s food-truck and small-vendor scene thrives on location, timing, and crowd flow. Yet most operators still rely on gut instinct or last week’s experience to choose spots. We wanted to bring data-driven insight—the kind big retailers use—to small businesses that move every day.

What it does

Foot Traffic Finder is a live, predictive map that shows how busy each area of San Francisco is now—and how busy it will be in the next few days.

Users can search any location to get a future forecast and a Gemini-generated summary explaining events and why it is rated that score.

The goal: help small businesses optimize staffing, inventory, and daily placement for maximum profit.

How we built it

Frontend: Next.js + TypeScript, styled with Tailwind CSS and Radix UI, animated via framer-motion.

Map: Mapbox GL JS for a full-screen interactive experience with dynamic data fetching (React Query).

Backend: FastAPI (Python) serving endpoints for weather, events, and traffic data.

Prediction Model: Multiple Linear Regression (scikit-learn) trained on historical “Popular Times,” weather forecasts (Open-Meteo), and event data (Google Events via SerpApi).

AI Summaries: Google Gemini API turns prediction data into human-readable insights.

Challenges we ran into

Handling API rate limits and combining asynchronous data sources (weather + events + traffic).

Accomplishments that we're proud of

Built a fully interactive, predictive map that is user-friendly! Utilized APIs by debugging over several cruel nights!

What we learned

How important small UX/UI touches are Best practices for utilizing the API calls without running into rate limits

What's next for Foot Traffic Finder

Integrate a more robust statistical prediction for future dates, linear regression is an OK start, but we could utilize machine learning models with confidence intervals to make our data even MORE accurate!