413 AI-Guided Skin Care
allows treatment of all the barriers to skin health, rather than focusing on only one skin
concern. Here are examples of BST nomenclature and treatment approach:
• DSNW (dry, sensitive, nonpigmented, wrinkle-prone): This skin type needs barrier
repair and soothing products to manage dryness and inflammation, while also
addressing anti-aging concerns. Which sensitive skin subtype this DSNW type has
(acne, rosacea, stinging, or allergic) further dictates which ingredients are appropriate
and which should be avoided. Acne-prone subtypes have very different needs than the
erythematous rosacea-prone subtypes.
• ORPT (oily, resistant, pigmented, tight): This requires products that control sebum
production and prevent and treat hyperpigmentation. The pigmented types will need
both treatment and maintenance regimens to manage hyperpigmentation as it comes
and goes.
BST CODING
Each BST is identified in two ways: a colored and numbered octagon known as the skin
type octagon (Figure 4) and a four-letter code.
For instance, BST 14: DRNT (dry, resistant, nonpigmented, tight) signifies a skin type
that is dry, resistant, evenly toned, and firm. It is represented by a light blue skin type
octagon with a 14. This coding system helps both consumers and skin care professionals
easily identify and remember the specific needs of their skin, while facilitating coding
for machine learning purposes. Data have shown that some people remember the four
letters, while others remember the number or color, so all three components are used when
Figure 4. These are the 16-skin type octagons that correspond to the BSTs. They are trademarked and
cannot be used without written permission to maintain the scientific integrity and proper use of the BSTS.

414 JOURNAL OF COSMETIC SCIENCE
possible. However, these designations are trademarked and must be licensed so that they
are used properly to prevent corruption of the system.
SKIN CARE ROUTINES TO TREAT SKIN TYPES
The skin typing system is much more intricate than it seems on the surface. AI has been
a giant leap forward in the ability to process this categorical information to quickly and
correctly provide the best dermatologist-recommended skin care advice. There are many
skin conditions and preferences to consider when developing a custom routine that goes
beyond the 16-skin type designations. For example, of the 16-skin types, 8 of them are
sensitive skin types that can have any combination of the 4 sensitive subtypes. This gives
a total of 136 different combinations of skin concerns that should be addressed with a
skin care routine. The degree of dryness and oiliness is also taken into consideration based
on the skin type quiz score, as is the presence of skin diseases such as eczema, seborrheic
dermatitis, allergic contact dermatitis, rosacea, and psoriasis. It takes more than 40,000
regimen combinations to treat all the varied skin needs. Each skin routine must have
ingredients that treat the underlying conditions, work well with other ingredients in
the routine, and do not contain ingredients that will worsen any of the underlying skin
conditions.
PRODUCT APPROVAL
Over the past 20 years, beginning at the University of Miami in 2004, thousands of products
of different formulations were tested directly on patients’ skin with the BST collected
in the dataset. Products that performed well for a particular skin type were tagged as
effective meanwhile, those that caused concerns, worsened skin, or did not have the desired
effect were labeled as not effective. Several dermatologists at the University of Miami
were involved in collecting these data over the years. Once thousands of products from
multiple brands had been studied, the products were compared to find similarities in the
effective and not effective products. This facilitated identification of which combinations
of ingredients, formula types, and product characteristics worked best on each skin type.
The first group of products that matched each BST were revealed in the book The Skin
Type Solution.25
In the following decade, dermatologists continued evaluating the products and combining
them together to determine which combinations of products were effective and did not
cause side effects on the 16-skin types and 4 sensitive subtypes. When side effects were
seen, they reported this information back, and the algorithms were adjusted. The 10 years
of testing by 100 dermatologists who used the system, in addition to another 100 doctors
and medical providers, were an invaluable part of the development of the algorithms.
Patterns emerged that led to more criteria that the products needed to meet to be able to
display a corresponding skin type octagon.
One criterion that was added later, which was not immediately apparent, was that the
brand’s marketing material for the product also was required to match the skin type.
For example, some products that claimed to be for dry skin would injure the skin barrier
in testing and were more appropriate for oily skin types. However, when these were
recommended to patients with oily skin, they would become confused when the product