Anabel Maldonado knows what she likes. A certain e-commerce site she shops at regularly, however, does not.

“I feel like I should just never see gingham or checks or floral print or denim jackets and things that I’ve never shown interest in,” she said. “Instead, I’m having to literally go and filter 20 brands that I like and a bunch of colours that I like … I’m trying to whittle down and curate on my own.”

In her view, that’s not how it should be. Maldonado is the founder and chief executive of Psykhe, a “personalisation-as-a-service” platform for retailers that tries to understand shoppers’ tastes and recommend products using artificial intelligence and psychographic profiles built from their behaviour. A customer who browses Rick Owens is more likely to exhibit what Maldonado calls “high openness, high neuroticism” than one perusing Tory Burch, so if they were to go explore fragrances later, the algorithm would understand there’s a greater chance they’d be drawn to dusky oud than sunny florals and surface those scents in real-time.

The start-up, which said its customers include Altuzarra, Farm Rio, Pacifica Beauty and Kirna Zabete, is trying a novel approach to a problem that’s plagued retailers since the advent of e-commerce: Without the benefit of a sales associate who can read a shopper and interact with them face-to-face, how do you help them find products they’ll like? That need becomes more pronounced as the retailer grows. Unconstrained by physical limitations like clothing racks and square footage, they can carry a vast inventory that offers shoppers endless options. But it also becomes overwhelming to sift through and diminishes any sense of curation, turning them into generic digital warehouses competing on price.

The problems inherent in this situation have entered the spotlight lately as retailers such as Farfetch and Matchesfashion have crumbled in recent months. Experts say their lack of curation was a primary cause, while the curation offered by retailers like Mytheresa, Moda Operandi and Ssense is a key factor in their survival.

A potential solution could be algorithmic personalisation, where the assortment a user sees is tailored to their individual tastes. The concept has been wildly successful in other industries. Spotify’s song-recommendation engine helped it become music’s dominant streaming service. TikTok’s rapid rise is due largely to its ability to hook users with its “For You” feed.

“You should be able to have a store that’s completely tailored to you — your taste, your size, your events, your lifestyle, the brands you like,” said Natalie Massenet, the luxury e-commerce pioneer who founded Net-a-Porter.

That store doesn’t yet exist. At many retailers, personalisation amounts to a carousel at the bottom of the page suggesting products that are visually similar to what a customer recently looked at or other items from brands they’ve searched. Even those ahead of the curve, such as Zalando, which asks shoppers their brand and size preferences to tailor its product recommendations and suggests items to go with past purchases, aren’t yet personalising at a level equivalent to something like Spotify.

The concept isn’t easy to implement in fashion, which deals with physical goods, not digital content. Still, it arguably presents a major opportunity — if it can be done right.

Knowing What Shoppers Want

Understanding what customers want and giving it to them is a simple idea but difficult to execute. To create its recommendations, Spotify starts to learn the relationships between songs by looking at which ones users frequently put in playlists together, it told the Wall Street Journal. It adds in metadata, such as the song’s release date and label, and runs audio analysis to rank characteristics such as danceability, acousticness, loudness, tempo, energy and mode, like whether it’s in a major or minor key. It examines the song’s lyrics, too, as well as adjectives used to describe the track in blog posts and articles online.

Spotify uses the information to build a multi-dimensional map of all the tracks in its library. Those positioned closer together are more closely related and thus probably appeal to the same listeners.

The technique isn’t unique to Spotify, but in fashion the method is tough to replicate. Part of what makes Spotify work is its huge song library, which has something for everyone. The fashion equivalent of songs would be inventory. But to maintain an inventory that size would be prohibitively expensive. It would also require cataloguing each product at the same level of granularity.

“Where I think it’s really hard is having all the data that you need to match the merchandise against the multi-dimensional space … of what the customer’s hyper-personalised preference would be,” said Holger Harreis, a senior partner at McKinsey and co-leader of its global data initiatives.

One company with its own version of the approach is Lyst, which acts like a search engine for fashion products offered by more than 17,000 retail partners. The company carries no inventory itself but charges a commission on sales through its site. It said it topped $600 million in gross merchandise value in its 2024 fiscal year. Because it effectively exists to help shoppers find what they want, it invests heavily in personalisation.

“At Lyst, we record and track absolutely everything that happens on our site,” said Anton Jefcoate, Lyst’s chief technology officer.

Each day it captures about 14.5 million data points, such as the amount of time a shopper spends on a page, how many times they scroll, the products they click on, what they put in their wishlist, the sizes they select and the colours they choose. Some are weighted more than others. Putting a product in a wishlist carries more value than scrolling, for instance.

The company gets product data from its partners but also uses a number of AI models to extract their characteristics. Similar to Spotify, it combines all this data into a kind of map, where each user and item is given a location in the form of a vector.

“It’s basically comparing your user vector against the vector of all the things in the database to find the closest match for a series of products,” said Jefcoate.

When shoppers browse Lyst, the products they see are tailored according to what Lyst understands about their tastes. The company conducted around 10 personalisation experiments in the past year and at best was able to boost conversion rates about 20 percent, according to Jefcoate.

Personalisation and Personalities

Whether this form of hyper-personalisation can work as well for every type of retailer or fully solve e-commerce’s curation conundrum is an open question. Massenet pointed out that shoppers look to higher authorities like editors and influencers for their cues on what to wear. Part of what makes some retailers successful isn’t just that they let shoppers easily find what they want but they also provide an educated opinion and a point of view. Otherwise you can wind up another nondescript outlet.

“You really need top-down curation with luxury products and fashion,” Massenet said.

Retailers also need to be mindful of how rigid their algorithms are. Psykhe’s Maldonado noted that surprise and novelty are important parts of shopping. If an algorithm is optimised strictly for conversion, its recommendations can be as stultifying and uninteresting as ones that aren’t relevant at all. In fact, Lyst has found weaving popular products into the results users see, even if they wouldn’t have appeared otherwise, to be effective.

“What customers really want — or at least what the numbers to tell us customers really want — is to have an element of personalisation in their product feeds but also be nudged and guided in the direction of other related but fresh start points,” Jefcoate said. “It’s not just about getting hyper-personalised. It’s about working out what the combination of ingredients is that really resonates with the customer.”

There may never be one Spotify of fashion. Not every Spotify user thinks its song recommendations are always on the mark anyway. But retailers are unlikely to go wrong by understanding what their customers want and trying to serve it to them. McKinsey’s Harreis said in his experience greater personalisation can increase conversion rates and reduce returns, which helps with profitability and has the added benefit of reducing the company’s carbon footprint.

Retailers just have to remember that sometimes what shoppers want is some guidance, too.

An Italian subsidiary of French luxury giant LVMH, Dior Manufactures SRL, on Monday became the third fashion company to be put under administration by a Milan court this year.

Prior to appointing special commissioners to oversee the Milan-based maker of Dior-branded handbags, the court had imposed the same terms on a unit of Italy’s Giorgio Armani group and on leather goods maker Alviero Martini, famous for its geographical map prints.

What Does the Court Administration Entail?

The decision to place a brand’s subsidiaries under court administration aims to give the companies time to reorganise their system of checks on external suppliers to ensure future compliance with existing labour laws.

What Does the Commissioner Do?

While the companies operate normally, the commissioner provides regular updates to the court on progress made in the setting up of checks, before magistrates decide to end the special administration.

What Are Magistrates in Milan Probing?

Milan prosecutors have for the past decade been investigating manpower supply companies that illegally employ workers, evading taxes and welfare and pension contributions to slash the cost of the services they supplied.

The probes traditionally targeted sectors such as logistics, transportation and cleaning services, where workers were supplied by firms that were opened and wound down every couple of years.

The focus has now shifted onto the fashion sector, where probes have highlighted similar problems this year as suppliers break employment rules to slash the cost of the goods they manufacture.

What Kind of Measure Is the Court Administration?

The prosecutors have been able to make use of a provision in the law that was originally designed to deal with companies infiltrated by the mob.

These companies would be placed under court, or judicial, administration through the appointment of special commissioners to oversee them.

By widening the scope of the provision to also tackle crimes relating to the illegal use of manpower, Milan prosecutors and police have been able to widen the scheme to include companies contracting goods or services supplied by third parties.

By Emilio Parodi and Valentina Za; Editors: Matt Scuffham and Jan Harvey

NAPA, CALIFORNIA – In launching Fara Homidi Beauty, celebrity makeup artist and founder Fara Homidi wanted to create products that are as efficacious to use as they are beautiful to look at.

“As a makeup artist, I have access to everything … I noticed that a lot of the industry standard best products out there weren’t clean,” she said, adding that those that did perform on clean standards were often lacking in shade range.

Fara Homidi Beauty, which was announced as the winner of the Creative Execution Award at the inaugural Business of Beauty Global Awards on Jun. 4 in Napa, California, brings a creative approach to designing professional-grade clean cosmetics. With products housed in sleek sky blue and deep orange packaging, featuring ergonomic shapes and luxurious touches, Fara Homidi Beauty also offers a makeup pochette made from vegan leather that looks more like a designer purse than a wash bag.

Beauty as Art

The idea of the beauty product as an objet d’art is commonplace, with no shortage of stylishly designed compacts, lipsticks and other cosmetics available for shoppers. However, many choose form over function, something that Homidi wanted to change.

“There’s a lot of great products out there…but I [am] maniacal about making sure the textures and ingredients were best in class,” said Homidi.

In her work as a makeup artist, Homidi has painted the faces of Hailey Bieber, Paloma Elsesser and Alexa Demie. Yet she was unable to find products that combined the long wear, colour payoff, shade diversity and superior textures she desired, alongside the clean credentials, such as being paraben, talc and sulphate-free, that she found important.

It took Homidi some years to find formulators and laboratories who could deliver to her exacting standards, while also being able to sustain the volume she needed. Saying she had to “push” to get the textures she wanted, Homidi said many of her requests seemed paradoxical.

“[I would hear] ‘you can’t have something that has glide and bounce that’s also velvety and powdery’. Those two things don’t make sense,” said Homidi. Her persistence paid off. The range, which includes lip liner, a complexion compact and a lip compact, is stocked in cult global retailers like Violet Grey in Los Angeles and Dover Street Parfums Market in Paris, whilst also being available in Mexico and Italy, and in Goop stores.

What’s Next

Fara Homidi Beauty may have begun its retail rollout with insider-favourite shops, but this month, it’s taking a major step into mass retail. On June 1, the brand became available to shop on Sephora’s e-commerce site, with a brick-and-mortar launch slated for 2025.

It’s a natural next step for Homidi, who was also accepted into Sephora’s Accelerate programme in 2022, which she described as “a six month crash course in all things business beauty.” The brand founder, who came to the U.S as an Afghani refugee as a child, believes that the time is right as the beauty world — and its customers — are ready for higher standards.

“I just felt like the time was now,” she said. “I’m someone who really follows my gut instinct.”

Next on her agenda is wider global expansion, including Australia, Canada and the UK once the brand is ready. New products that blend skincare and makeup are also in development.

The brand is open to all, says Homidi.

“That could be men or women … however you see your gender,” she said. “It’s for anyone who wants something that feels special that they can still use every day.