Recommend by scent with Smelling Similar

Have you ever visited a new place for the first time and been flooded with memories? Maybe it's of visiting your grandparents' kitchen or sitting on the couch watching cartoons after school. Scent memories are some of our strongest as humans. That’s why Algolia is proud to announce our latest machine learning model for Recommend — Smelling Similar.

We use scent-mapping algorithms designed by the perfume industry and a custom-trained machine-learning model to identify odor vectors between smells. Categorizing top notes, heart notes, and base notes allows the model to associate the interior smell of a new sports car with a baseball mitt or freshly baked cookies with a holiday party. You provide a list of existing smells, and Smelling Similar learns to identify their relationships to offer recommendations.

“We help our buyers think of every detail when finding a house,” says Barbera Coretta, CEO of Your4EvahHome. “Incorporating Smelling Similar into our listings page allows home buyers to ensure the place where they build their new memories smells like places where they’ve had good memories before.”

Smelling Similar provides site owners with significant control over recommendations, empowering them to establish thresholds for the "similarity" of smell matches and to filter out “bad” smells across your catalog.