trending engine

Case 2

Trending Engine

Inspiration

Most “trending” feeds feel like black boxes, you see results, but not why something is trending. We wanted to build a system that doesn’t just rank content, but also explains the underlying patterns using real statistics.

How we built it

We designed a lightweight backend using FastAPI with SQLite to handle posts, engagement signals, and scoring logic. The frontend is a clean vanilla JS dashboard that renders leaderboards, sorting, and a stats tab with distribution insights and charts. A decayed scoring function ensures recency is factored into rankings:

$$ score = (likes + 2 \cdot shares + 1.5 \cdot saves) \cdot e^{-\lambda t} $$

What we learned

How to design a ranking system that balances engagement and recency
Translating raw data into meaningful insights (median vs mean, variance, outliers)
Building a clean separation between API, data layer, and UI

Challenges

Avoiding naive implementations (like constant polling and inefficient queries)
Handling skewed data where a few viral posts dominate metrics
Turning raw statistics into intuitive, useful visualizations instead of just numbers

Outcome

The result is not just a leaderboard, but a mini analytics engine that reveals how content actually performs, highlighting consistency, volatility, and true “trendiness.”