AI Case Study

Proven offers personalised skincare products with the use of machine learning

Proven leverages machine learning and semantic search to offer its customers personalised skincare products. With a database of more than 8 million consumer reviews on more than 100,000 existing beauty products and 20,000 beauty ingredients the system is able to make connections between products, ingredients and reviews and create categories for different skin types, ethnic background and geographic region. Then, when a consumer completes a short quiz about their skin, they get a matching skin care regimen with custom-made products to fir their unique skin profile.

Industry

Consumer Goods And Services

Personal And Household Goods

Project Overview

"Ming Zhao and her co-founder, Stanford scientist Amy Yuan, technically founded Proven a little under a year ago, though the database at the center of its business model was about two years in the making. 

“This database encompasses more than 8 million consumer reviews about skincare products, more than 100,000 beauty products that are on the market and 20,000 beauty ingredients and more than 4,000 scientific articles or peer-reviewed journal articles about skin and about ingredients,” Zhao explained, noting that the database allows for transparency about suppliers, ingredient origins and efficacy.

Bots continuously scrape the data in the database and, through machine learning, are able to make connections between different product categories, ingredients and review ratings. Zhao explained that Yuan would use semantic searches to pick out different key words ― acne, wrinkles, etc. ― and then make sense of the ingredients and how they affect those skin concerns for different people. All of that data is then further filtered into categories relating to skin type, ethnic background and geographic region.

On the user experience end, consumers just fill out a short quiz (developed by a dermatologist) that asks for things like age, skin type, skin goals, ethnicity and geographic location. Data from the quiz are then calculated (based on information from the database), and the user is presented with a unique skin profile and a suggested skin care regimen with custom-made products to suit their individual needs. "

Reported Results

Personalised products

Technology

Analysis of the database using machine learning to make connections between product categories, ingredients and review ratings and use of semantic searches to understand the ingredients in products and their effects.

Function

Marketing

Customer Management

Background

"People have been using beauty products to enhance their eyes, brighten their skin or smooth their hair since ancient Egyptian times. But over the years, the beauty market has grown into a $445 billion industry, with companies competing to sell us different versions of the same thing.
More and more companies are embracing the individuality of their customers, creating products designed specifically for each of them. With the help of new technologies, including artificial intelligence and machine learning, the possibilities seem infinite."

Benefits

Data

Database of more than 8 million consumer reviews about skincare products, more than 100,000 beauty products that are on the market and 20,000 beauty ingredients and more than 4,000 scientific articles or peer-reviewed journal articles about skin and about ingredients