NARABU - Queue Wait Time Prediction App

This is an app that uses AI to predict queues and wait times at popular restaurants, displaying them on a map with green, yellow, and red pins. The number on each pin directly indicates the predicted wait time in minutes.

Even for restaurants without reports yet, it provides a daily estimate based on genre, time of day, day of the week, and weather. People waiting in line can report in 10 seconds (2 taps, no text input), and as reports accumulate, the predictions converge with actual measurements.

What I was most particular about was not hiding uncertainty. Every prediction includes a confidence bar, and when the basis is thin, it honestly displays "AI Prediction (Reference)." While it would look better to present numbers confidently, once a prediction app makes a major miss, trust doesn't come back.

We don't use generative AI. We multiply genre-specific baselines by coefficients for time of day, day of the week, weather, and holidays, then blend them with weighted median values and freshness of recent actual reports.

Tech stack: Flutter / Firebase (Firestore, Cloud Functions, FCM) / Google Maps SDK / Places API / RevenueCat / AdMob. All crawling and prediction is handled entirely on the Cloud Functions side, and API keys are not exposed to the client.

Coverage starts from central Tokyo. We'll expand as more people use it.