The opening

Demand for personalized products keeps rising, but the customization process itself can be overwhelming, and the fragrance industry has only adopted basic online customization so far. Olfactory NYC already sells custom fragrance creation and wants it to be accessible to everyone, so the opportunity was to make the existing process easier rather than to invent a new product.

The assistant

SmartScent AI uses machine learning to suggest base, middle, and top notes from a customer's brand name preferences, their past creations, and current trends. It was designed to run on the website and in store on a tablet, so the same experience carries across both and a human associate stays in the loop for the consultation, which is a large part of what the business is actually selling.

Two demos, not a description

The front end demo walks the full customer journey, from entering scent preferences through a six step creation flow to confirmation. The back end demo connects the assistant to ChatGPT-4 and a perfume database and answers requests written the way people actually talk, like asking for something for a date who likes the beach.

Rollout and measurement

We set out how it would launch: initial development in months one and two, pilot testing through month six, evaluation and chain wide rollout in months six and seven, then full development and continuous improvement after that. For measurement we proposed customer satisfaction survey results and customer journey completion percentage as leading indicators, with overall perfume sales as the North Star metric, so the assistant gets judged on whether it sells more perfume rather than on how many people poke at it.