Nerd Meets Fashion Maven: Google's Boutiques.com

Google's boutiques.com experiment, launched in late 2010, was an unusually candid attempt by the company to enter the fashion discovery market by doing something Google almost never does: acknowledging that it doesn't understand the problem and building in a mechanism for humans to provide the taste that its algorithms could not generate.

The platform asked fashion experts — stylists, magazine editors, bloggers — to curate collections of clothing and accessories from around the web, creating personalized boutiques that visitors could browse based on whose taste they shared. Google then used machine learning to analyze users' browsing behavior and match them to the curators whose taste their behavior most resembled.

The admission embedded in this design was significant. Google's standard approach to any product problem is to build an algorithm that scales infinitely and requires minimal human curation. For fashion discovery, the company concluded that human taste was genuinely irreplaceable at the top of the funnel — that no algorithm, given only behavioral data, could reliably distinguish between objects that were beautiful and objects that were merely popular.

The fashion industry received boutiques.com with polite interest and some skepticism. The tension between Google's data-driven culture and the intuitive, hard-to-articulate logic of fashion taste was real and visible in the product. The curators who participated found the platform technically impressive but questioned whether its users would form relationships with fashion content through the Google interface with the same depth they formed them through dedicated fashion media.

The problem it tried to solve — connecting people to fashion they would love — remains one that both algorithmic and human approaches have addressed imperfectly.

[Editor's note: Boutiques.com was shut down in 2012, its technology folded into Google's broader shopping products.]