The science of knowing your taste

The cold start problem

Every recommendation system faces the same challenge: how do you recommend things to a new user with zero ratings? CatNLog solves this with import — bring your Letterboxd, Spotify, or Goodreads history and skip straight to personalized results. Even 20 ratings are enough to build a meaningful taste profile.

Collaborative filtering

The core insight is simple: people who agreed in the past will likely agree in the future. If you and another user both rated 50 films similarly, the films they loved that you haven't seen are strong candidates for your next watch.

CatNLog uses matrix factorization techniques (ALS and SVD) to decompose the massive user-item rating matrix into latent factors. These factors capture hidden patterns — you might not know why you keep rating certain films highly, but the math does.

Content-based filtering

This approach analyzes the attributes of items you've rated highly — genre, themes, tone, complexity, era, creator — and finds other items with similar attributes. It's particularly useful for niche tastes where few other users share your exact preferences.

Hybrid approach

CatNLog combines both approaches. Collaborative filtering catches items that your taste-twins love. Content-based filtering catches items that match the attributes you prefer. Together, they deliver recommendations that feel both surprising and right.

Cross-category intelligence

What makes CatNLog unique is that your taste profile spans all four categories. Love atmospheric horror films? You might also enjoy atmospheric horror games, dark ambient music, and gothic novels. The AI sees these connections across media boundaries.