Home - T Singapore

Would You Let Artificial Intelligence Select Your Skincare?

By Bianca Husodo

Felicia Yap

In the olden days, you could walk down the street to mini apothecaries and neighbourhood drug stores, saunter in while encountering familiar faces and proceed to inform an in-store chemist of your current skin ailments. The chemist — who you might already be acquaintanced with, hence armed with first-hand of your exact needs — would proceed to compound the products then and there.

The contemporary beauty industry is nothing like that. Mega beauty stores mushroom across shopping malls, where an ocean of products lies within. Swarming beauty advisors snake between the influx, eager to thrust the latest and hottest to the clueless and confused. Dodge their hovering presence and it still doesn’t get better. Tubes, jars, bottles line shelf after shelf. Slapped on them are labels spelling out lists of ingredients foreign to the tongue. All these sealed with seductive promises of rejuvenation and transformation for the skin.

A one-size-fits-all system formulated for the demographic masses fits no one, according to a new class of beauty start-ups powered by artificial intelligence (AI).

The AI Consultant

During her college years, Siqi Mou — now co-founder and CEO of HelloAva, an AI-fuelled skincare personalisation startup — hesitated to raise her hand in class. The Harvard and Standford Business School graduate was wary of the heat of scrutinous eyes that would bore into her acne-ridden visage. The condition swelled into self-consciousness and insecurity. Mou would ask her peers for advice and they’d suggest several topical solutions. What worked wonders for them, none worked for her. Simply put, their skins were different.

“Almost everyone has to go through trial and error to know what works. I wanted to create a solution in the beauty industry,” said Mou in a phone call. “I thought if we could lead people to the right products without the trial and error process — but also share their trial and error experience to the community — that would be a huge improvement on the current status quo.”

Wenjun LiangSiqi Mou, the forward-thinking 29-year-old behind HelloAva.
Siqi Mou, the forward-thinking 29-year-old behind HelloAva.

Mou, who has her background in finance, switched gears to beauty when an epiphany hit. Inspired by the success of dating apps — of which algorithms can match people with common interests far more accurate than humans could — Mou figured there were unexplored potentials for the same logic to be applied to skincare and beauty.

“Humans can be biased and can miss certain data points,” she reasoned. “Everyone’s skin is different, and we’re at a point where technology enables us to do so much but it just hasn’t been applied in the beauty sector — it’s so behind and antiquated.”

Launched in 2017, HelloAva eliminates the guesswork out of the traditional beauty experience. Driven by data, it acts as a personal bot-cum-skin consultant that curates products tailored to the specific needs of one’s skin type. I decided to give it a spin.

“What are your primary skin concern(s)?,” Ava, the chatbot, inquired.

Underneath her text bubble, ten options popped up, offering “Acne” to “Not Sure” as potential answers. Decisively, I selected “Dark Spots”, “Oiliness” and “Clogged Pores”.

Next, Ava asked for the types of beauty brands I liked, before moving on to my preferred products, whether I wanted them as part of a 3- or 5-step regimen, or perhaps customised to suit my own preference. My cursor hovered, scanning what each entailed. The 5-step regimen — cleanser, exfoliator, toner, moisturiser, SPF — sounded ample without being excessively rigorous. I clicked.

Meet Ava, HelloAva’s chatbot.
Meet Ava, HelloAva’s chatbot.

Ava then requested an optional photo of my face, “to better identify your skin conditions,” she explained. I snapped a selfie, attached it and Ava continued down her line of questions, her queries delving deeper into the state of my skin. In total were about 30 questions, succinct and straightforward, taking about five minutes to complete.

Mou named her chatbot after ‘Ex Machina’’s humanoid robot. “She’s very smart and she becomes smarter,” Mou said, juxtaposing Alicia Vikander’s fictional character to HelloAva’s machine-learning algorithm.

After one completes the questionnaire, the data points gleaned from it are assessed by a self-learning algorithm, which mimics how dermatologists sift one’s skin into a typology and suggest products.

Following the skin profile quiz, a report on my skin was generated. OSPT, spelled its headline — an acronym for oily, sensitive, pigmented and tight. Below, a summary of said skin typology, weighing in on the pros (“... resists ageing better than many other types.”) and cons (“... oiliness can lead to acne.”). This is one of 32 skin types that Mou and her team identified in consultation with dermatologists. The product recommendations weren’t issued right after, requiring a human skincare advisor’s verification before the products were finalised.

“We take inputs from our experts and allow them to modify products. In AI terms, we call this process the supervised learning process. Meaning: Your recommendation result is being modified by a human. Every single time the human modifies, it allows the algorithm to learn from a human touchpoint and become slightly smarter over time,” Mou explained.

These human inputs aren’t just sourced from HelloAva’s skin experts. “The algorithm takes in feedback from people after they use the product,” said Mou. In turn, the collected feedback refines the algorithm, allowing it to hone its recommendation accuracy.

Each recommended product features a matching score, anchored in its compatibility level to a skin typology.
Each recommended product features a matching score, anchored in its compatibility level to a skin typology.

Within three days, a client can view and modify the curated products. On the recommendation page, the products are segmented according to their types — cleanser, toner and so on — each including up to five different brands. These are taken from the database’s stable of brands, which pools in more than 80 names, from time-tested prestige brands to cult Korean indies.

When I received my personalised product curation, I was eerily taken aback. Most of the products I’ve already owned and learned to have worked for my skin, through prolonged years of hits and misses, were in the curation. If only HelloAva existed earlier, I thought.

Taking AI Personal

Pushing it one step further is Proven. Masterminded by entrepreneur Ming Zhao and Stanford University computational physicist Amy Yuan, Proven similarly taps into the potency of AI for the ultimate bespoke beauty experience. But it doesn’t stop at curation. The brand concocts products from scratch, custom-tuning it down to the ingredients.

Like Mou, both Zhao and Yuan were driven by their personal skin gripes to start Proven. For Zhao, she underwent what felt like an insurmountable quest for the right skincare products, before succumbing to a visit to a skin guru who tailored products specifically for her. “It made so much sense: Something that is made with you and your needs in mind, will work better,” she said.

ProvenProven’s personalised products (left) and Ming Zhao, co-founder and CEO of Proven (right).
Proven’s personalised products (left) and Ming Zhao, co-founder and CEO of Proven (right).

While they were exactly what Zhao was looking for, they came with a hefty price tag. It was only after she met Yuan — who self-built the Skin Genome Project, her own AI-powered database, to battle her raging atopic dermatitis — that everything fell into place. Harnessing on AI, Yuan gathered and combed through thousands of scientific journals, governmental and clinical research papers on her condition. Combining that with consumer testimonials, reviews and conversations surrounding it, Yuan filtered them with natural language processing and other machine-learning algorithms to extract the consummate solution.

“We were intrigued by each other’s research and how we solved our own issues. For me, it was personalised products. For her, it was data-powered knowledge,” Zhao said. “We combined these two epiphanies and created Proven.”

Furthering the Skin Genome Project to address all skin concerns, it’s currently the largest database of its kind, claimed Zhao. It rallies more than 4,000 scientific journals and 20 million consumer reviews. The pair consulted Dr. Tyler Hollmig, Head of Aesthetic Dermatology at Stanford University, to funnel down the database’s suggested ingredients to an underlying formulation that made sense.

“Our goal with the Skin Genome Project is for it to become the de facto proof for beauty. As the database grows, the accuracy of it is improved. You no longer have to ask a random person who may have a different skin as you what they’re using that works.”

Similar to HelloAva, clients start with a skin assessment which will chart 27 different factors — from skin tone to level of air pollution. The result will group them into different skin typologies, which will lead to the next step: creating a set of dedicated products to fit the typology. There are more than 200 permutations of products, according to Zhao.

ProvenDubbed the CPR System — Cleanse, Protect, Repair — Proven’s three-step skincare kit comprises of a daily cleanser, an anti-pollution sunscreen and a restorative night cream.
Dubbed the CPR System — Cleanse, Protect, Repair — Proven’s three-step skincare kit comprises of a daily cleanser, an anti-pollution sunscreen and a restorative night cream.

The Proven “box set” — a simple three-step regimen of a daily cleanser, an anti-pollution sunscreen and a restorative night cream that costs S$265 (US$195) — is guaranteed to treat every skin type and skin concern with visible results within a few weeks of daily application.

“55 percent of people are dissatisfied with the beauty purchase they make post-purchase. They may have done their desktop research, consulted with experts, magazines — and yet after all of this personal R&D, they are most likely unhappy with the acquired products. This is the industry where trial and error is fuelling a lot of the revenues of the industry overall. And there’s little incentive for the companies to change that,” Zhao stated.

At the moment, both Proven and HelloAva only cater to the United States and Canada. Yet as these start-ups gain more momentum — HelloAva served 40,000 users in 2018 alone; Proven counts 1,000 clients since its product launch a month ago in December — they are likely to spearhead a wave of tech-beauty revolution.

“Somehow skincare and beauty — which for a large part of the population are very important, intimate daily consideration — have completely escaped the attention of technology. It hasn’t seen much improvement for the past 50 years, and that’s despicable,” Zhao declared. But a changing of tides is approaching: Zhao, Yuan, Mou, along with other nascent tech-beauty groundbreakers, are shifting the course towards a better consumer experience.

“With the advent of AI and technology, it’s kind of funny that we’re coming back to this mode of being able to serve customers as if we know you,” noted Zhao on how the industry is coming full circle to the personalised experience of past-century apothecaries.

“Putting the personal into personalised, that’s what we believe will be the future of skincare,” she said. “Or actually, the future of all consumer products.”