The Data Science Behind our Products - IngredientRank™

The Data Science Behind our Products - IngredientRank™

There are hundreds of ingredients that we could use in our products. Some are proven to be more efficient than others, some work better with other ingredients, others have dubious safety or efficacy records.

At UXB we wanted to take the luck out of finding the best ingredients for our range by developing an algorithm called IngredientRank™.

IngredientRank™ was developed to find the best ingredients (and ingredient combinations) for a particular skin condition. We have worked with top data scientists and cosmetic formulators to tune the algorithm to get it to a place where we can reliably ask it to provide the top ingredients for mature oily skin, for example.

Here’s a high level view of how it works. To start we feed three things into the algorithm:

  • A corpus of scientific papers, blog posts from domain experts and ingredient lists from existing products.
  • A database of ingredients that we know we can sustainably and ethically sourced (including INCI, common names, abbreviations and synonyms).
  • The skin condition or type that we are looking to help.


The algorithm uses advanced text analytics and a bespoke scoring algorithm to produce a ranked list of ingredients for consideration. The ingredients at the top of the list have co-occurred in the corpus more frequently with the targeted skin condition. This means they are being studied to treat this condition or they are actually being used to treat the skin type.

Ingredients that are used together in formulations get linked and their rank increased. Some ingredients ,ferulic acid and vitamin C for example, support or enhance the other ingredients efficacy. The output is a kind of weighted graph. Here's a made-up example for a search on ingredients for oily skin.

ingredientrank


The more certain the algorithm is that the ingredient is beneficial to this condition, the larger the node. Some of the ingredients will be mentioned in a negative way in relation to the skin condition. We use semantic detection to work out if this ingredient is good or will have detrimental effects. In the visualisation above these ingredients have a different colour. Gold is good. Black is bad.

IngredientRank™ gives us our superset of ingredients. It takes the human bias out of ingredients selection but it is then down to our cosmetic chemist to decide which ingredients we use in which products. Formulating a product is as much an art as it is a science. We are designing for the job to be done (cleanse, moisturise, exfoliate etc.) but also for skin feel, viscosity and ease of application.

The tagline of UXB is ‘The application of nature™’. This tagline has many meanings for us but chiefly it encapsulated our desire to apply nature to specific skin concerns. We have the utmost respect for herbalists and natural formulators who develop products that use mineral and organic ingredients to support the skin. Our IngredientRank™ algorithm is but a tool that builds on the millennia of wisdom and academic research that has gone before.