Source: Business of Fashion, May 2012
… in the complex market for fashion, with its subjective tastes, trend cycles and gatekeepers, many debate whether Netflix-style recommendation algorithms are the answer.
The failure, last September, of Google’s Boutiques.com — a fashion site that promised perfectly personalised product selections powered by “machine learning” — seemed to sound something of a death knell for purely algorithmic recommendations in fashion. And yet, despite the spectacular rise of social curation sites like Pinterest, which features thousands of fashion products hand-picked by humans, simply presenting users with items shared by the people they follow is also an imperfect solution to the personal relevance problem.
“Algorithms and search functions just do not account for taste and style,” said Suraj Kapoor, co-founder of LookLab, an angel-funded style advice site that enables people to ask questions pertaining to specific purchasing decisions and get personalised style advice from “fashion insiders” who range from junior stylists to bloggers and are free to suggest products from any e-commerce site. When posing questions, users can also indicate their personal style, favourite brands, celebrity inspirations, body shape and budget. “We allow users to communicate one-on-one with a fashion industry insider for personalised style advice. It’s like having your own free stylist, online,” Kapoor continued.
A growing number of fashion sites at the affordable end of the market now promise consumers personalised product selections, picked by Hollywood-type “celebrity stylists,” in response to short style quizzes of 15 to 20 questions that aim to understand a user’s personal style and product preferences. London-based Stylistpick, which has raised a total of $19 million from investors including Accel Partners, Index Ventures and Fidelity Growth Partners Europe, promises consumers a “personalised showroom” full of products selected by “A-list stylists” including the host of MTV makeover show “Plain Jane.” Both monthly subscription and pay-as-you-go options are available.
JustFab, which, last Autumn, raised $33 million in Series A funding led by Gilt Groupe-investor Matrix Partners and has attracted over 6 million members, provides users with product selections from a team of expert style advisors led by celebrity stylist Jessica Paster. Sites like SoleSociety and Send the Trend operate in a similar fashion, offering personalised style selections curated by fashion experts in response to a style quiz. But many of the so-called experts who are curating commerce and offering advice on these sites are relatively unknown and have little credibility within the fashion community.
… angel-funded fashion site Feyt aims to bring greater fashion knowledge and industry credibility to the personalised e-commerce space. Founded by Eleanor Ylvisaker, who co-founded Earnest Sewn, and Ferebee Taube, who was previously a consultant for clients including Vogue, Theory and Alexander Wang, the site — which targets the luxury end of the market and features brands like Lanvin, Jil Sander and Proenza Schouler — will blend expert curation with algorithmic recommendation to deliver personalised styling recommendations that point users to products on external retailer sites.
“Our focus is shifting shoppers from search-based purchases to suggestion-based purchases,” said Ylvisaker. “When the site launches in full we will combine algorithmic technology with real style expertise to offer customised shopping recommendations for each individual user. We think of ourselves as the next generation of luxury e-commerce.”
Unlike many other curated commerce sites, Feyt’s vision is less about the appeal of receiving personal recommendations from a stylist and more about the actual quality of those recommendations. “The most important role our fashion insiders play is on the back end, in order to execute a scalable product,” Ylvisaker continued. “We developed heuristic guidelines based on style preference, body type, lifestyle and seasonal trend,” she added. “Every fashion item shown on Feyt.com gets tagged based on these guidelines, which in turns feeds the algorithm to provide our members with personalised recommendations that match their profiles. We’re not just another blog showcasing our own personal style in the hopes of inspiring you. We’re showcasing your personal style in the hope of inspiring you.” In addition to personalised purchase proposals and styling advice, the full site will include features like virtual wardrobe planning and suitcase suggestions.
But what makes Ylvisaker think that Feyt — which plans to make money through affiliate commissions and sponsorships and is looking at additional monetisation via white-labeling and data analytics — will succeed where Google’s Boutiques.com failed?
“We have found that other fashion sites that have attempted to use algorithmic technology are lacking in the fashion understanding and styling background that we can provide,” said Ylvisaker. “Plus, the product itself is selected by us, so we are curating the data set from the start.”
… companies like Feyt must surmount some significant challenges. Adequate data is perhaps the biggest. The more user data a recommendations engine has to work with, the better its chances of producing useful recommendations. Next month, in advance of the full launch this fall, Feyt will launch a virtual closet feature, which will allow users to create and share looks with a network of friends, as well as generate data that will help feed the site’s recommendation engine. Whether this will suffice remains to be seen. It’s no coincidence that companies with some of the most successful recommendations systems, like Netflix, also have significant volumes of user data. Dealing with constantly changing data is another major challenge for any algorithmic recommendations system in fashion, a market that’s increasingly driven by ever-faster trend cycles.
But while a purely algorithmic play may be a challenging proposition in fashion e-commerce, intelligent algorithms can undoubtedly help a human stylist or other expert be more efficient and deliver personalised recommendations at scale. Indeed, for the foreseeable future, the ultimate solution to that perennial problem — what am I going to wear? — is most likely to be a blend of man and machine.