Creating a new fragrance formulation is an iterative and time-intensive process. Perfumers begin with a target scent profile - selecting top, middle, and base notes - but translating that creative intent into a stable, manufacturable formula requires extensive trial and adjustment. Ingredient interactions, concentration ratios, and stability under varying conditions all affect the final result in ways that are difficult to predict without repeated experimentation.
The team had accumulated a significant database of historical formulas over the years - but this knowledge was not being leveraged systematically. When a perfumer started a new formulation, they relied primarily on personal experience and intuition. There was no mechanism to surface relevant past formulas, suggest compatible ingredient combinations, or flag ratio ranges that had worked for similar scent profiles - forcing every new creation to largely start from scratch.
Problem Statement
THE CHALLENGE
Challenges
-
Long development cycles
3-9 months average time to develop and validate a new fragrance formula from initial brief to approved formulation.
-
High number of trial iterations
15-40 trial formulations required per recipe before achieving the desired scent profile and stability specifications.
-
Existing formula database not being utilised
Years of successful formulations sat in a digital database with no intelligent layer to surface relevant combinations or patterns for new briefs.
-
Heavy reliance on senior perfumer expertise
New formulation decisions depended entirely on individual experience, creating knowledge bottlenecks and inconsistency when senior perfumers were unavailable.
Solution
THE SOLUTION
We built an AI-powered recommendation engine trained on the company's existing digital formula database. The perfumer inputs their selected fragrance notes and target accord - and the system recommends compatible ingredient combinations and ratio ranges drawn from patterns in historical successful formulations. Rather than starting from a blank slate, the perfumer begins with a ranked set of AI-recommended starting points grounded in what has worked before.
-
Recommendation engine trained on existing digital formula database - no digitization or data capture required.
-
Perfumer inputs selected fragrance notes and accord direction to receive ingredient combination suggestions ranked by compatibility.
-
Ratio range recommendations surfaced from patterns in historical formulations with similar scent profiles.
-
Dashboard displays ranked suggestions with ingredient compatibility scores, allowing perfumers to compare and select starting points.
-
System learns continuously as new approved formulations are added to the database over time.
