A Fashionista Hal

The September 10, 2012 issue of The New Yorker has a profile of Federico Marchetti (in the World of Business column by John Seabrook - - the article is entitled The Geek of Chic).  Marchetti is the founder and CEO of the Yoox, a leader in the world of e-commerce for fashion.

What is interesting is the role of predictive algorithms in the fashion business.  From the article:

During my visit to corporate headquarters, I met the keeper of Yoox’s algorithm, Alberto Grignolo, a husky, balding man with pale eyes and a large smiling face.  He explained that the algorithm, which is made up of some four hundred pages of code, is used by Yoox’x buyers the help them decide how much of a particular item to order and the target price.  “The usual question they ask is ‘Tell me what will sell.’  So we designed a set of rules – if you are generating a lot of profit margin with a particular product, then that is the product you should buy more of, because you will make more money.  But don’t buy much more, or you will have too much left over.”  Of course, bricks-and-mortar buyers make the same decisions, but they tend to be far less data-driven, and guided more by intuition.  That they are wrong a fair amount of the time is the reason so much overstock exists.

“But it’s a weird world, fashion,” Grignolo went on, “because there are swigs in demand fueled by trends, by designers, by magazines telling you what to wear next season.  So you know a certain item is profitable, but to what extent will it be popular next season?  The algorithm knows that last season two per cent of our sales were in pink items, let’s say, so next season do two percent pink again, whereas a buyer may think pink is going to be hot – because the trendsetters will make it hot – so he buys five per cent.

The algorithm also sets prices on the discounted items.  In the course of a year it will set a hundred million prices – far more than a bricks-and mortar store could manage.  For the customers, the algorithm tries to serve as “the perfect clerk,” as one Yoox worker put it to me, by showing you items based on you purchase history and on the purchase history of others who have bought things you’ve bought.  However, these kinds of recommendations can be trickier in clothing choices than in, say, books, because people don’t wear what other people are wearing.  People dress to look different.

As was often the case when I spoke to people at Yoox, Grignolo gave me a glimpse of the future.  In linking to the two planets, fashion and the Internet, Marchetti has unleashed forces that my eventually disrupt everything about the fashion business.  Using the Internet to buy clothes is only the beginning; in the future, buyers might use the kinds of funds of data that Yoox mines to make better-informed purchase decisions, which would reduce end-of-season overstock.  Future trends might be predicted by an algorithm – a sort of fashionista Hal.

Grignolo didn’t think that would happen.  “The algorithm predicts things based on what it has already seen.  You, the buyer, think this designer is going to be hot his season – well, the algorithm can estimate the likelihood that you are completely wrong.  But it will not necessarily be able to predict the next hot designer itself.”