How Unilever is ensuring inclusivity in its AI-powered beauty experiences

Online shopping has become a fixture of the global retail market, with the total share of online retail sales surpassing traditional in-store sales in the US for the first time in 2019.

Though one-click buying is undeniably convenient, at least one helpful element of the in-store experience is missing — a store assistant.

In stores worldwide, staff are on hand to answer a simple question: “What do you think would work best for me?”

This is doubly true when it comes to buying skincare or haircare products. Though there are countless articles online about the “best moisturiser for dry skin” packed with ad placements, it takes a long process of trial and error to find the right one for your skin type.

To solve this, consumer goods giant Unilever developed its own AI beauty assistants in-house: BeautyHub Pro and Pond’s Skin Institute.

Available in the Philippines and Thailand, BeautyHub Pro is an AI-powered selfie tool that offers product advice for skincare and haircare on a downloadable app.

By completing a quiz and sharing a selfie, BeautyHub Pro uses its Computer Vision AI to ‘see’ and assess up to 30 visual data points and offer personalised product recommendations from one of Unilever’s brands, such as Dove, Dermalogica, Paula’s Choice, TRESemme, or Vaseline.

Similarly, Pond’s Skin Institute, available in the Philippines and Indonesia, uses users’ selfies to identify skin care concerns and recommends a product from the Pond’s Skin Institute range.

Aaron Rajan, global VP for consumer experience technology at Unilever, says these offerings follow its Beauty & Wellbeing Business Group’s mantra: “Purpose, science, and desire.”

“This means creating purposeful and meaningful brands that positively impact people and the planet, using cutting-edge science and technology to deliver unmissably superior products, and continuing to make our brands relevant and timeless,” Rajan explains.

Getting under the skin

Rajan says the foundation of BeautyHub Pro and Pond’s Skin Institute lies in extensive R&D clinical studies spearheaded by Unilever.

He explains that the team used algorithms trained on a diverse dataset encompassing all skin tones and hair types.

“Data and AI are intrinsically linked,” says Rajan. “Data gives us the power to make informed decisions and drive action based on business insights, while also providing the fuel.

“AI is the engine that transforms raw information into actionable insights and content.”

He adds that the quality of an AI model heavily depends on the quality and relevance of the data training it. So, a thoughtful approach to data collection, preparation, and usage directly impacts the effectiveness of AI systems.

“It’s important that we first develop a solid data foundation. For these personalised beauty experiences, this could be consumer research to better understand our consumers or a knowledge base of information about the products.”

The team will then test the AI before deployment to ensure the quality of the responses.

“We train the model using a mix of deidentified data from various skin and hair types to cover a wide range of needs,” Rajan says.

“The AI then analyses aspects such as hydration, redness, wrinkles, frizziness, volume and colour consistency.”

Quality testing

The team will continually update the AI with new data and feedback to keep its recommendations accurate and effective.

For instance, to ensure the accuracy of results for its algorithm for hair damage, its tool has been trained on a dataset of over 10,000 standardised selfies – gathered with full consent as part of its research and development, and lab images across a diverse range of hair types.

It certified that each image was standardised in style, such as similar head positions and lighting. Then, it was tested on over 150,000 additional data points to check the quality of the responses provided.

To ensure security, Rajan explains: “Unilever reviews all AI-powered tools before deployment with a cross-functional team of experts including privacy, cyber security, and local legal teams to ensure compliance with applicable laws and best practices.”

“Given the speed at which AI is developing, we have faced challenges in moving into this space where there are so many unknowns,” says Rajan. “Working collaboratively with our tech partners, internal tech teams and brands, we have learned together and developed these powerful experiences for our customers.”

Selfie-assessment

Taking a selfie and receiving a product recommendation saves users time and money spent on trialling and testing multiple products.

“Consumers have shared the feedback that the BeautyHub Pro quiz is simple and easy to use, is credible with expert advice, and provides helpful product recommendations,” says Rajan.

He explains that the feedback is reflected in its data. For BeautyHub Pro, a basket size is almost 40% higher than those who shop elsewhere, and consumers are 43% more likely to complete a purchase than those who browse on other channels.

In 2023, more than 80,000 users experienced BeautyHub Pro, and it’s expected to reach more than 3 million people.

Whereas for Pond’s Skin Institute, more than 30,000 consumers have used its AI Skin Expert, and the brand is predicting more than 100,000 users by the end of this year.

For the future, Rajan explains that Unilever is testing how it can use technology and AI to continue providing personalised and immersive experiences.

“We see AI continuing to play an important role in helping consumers learn about their skin and hair concerns and recommend the right products and regimes.

“AI-assisted beauty is something that will stay but will take different shapes and forms.”