some small online shoe vendor at the other. A shoes website, for example, might have all their products categorized as shoes, even though some of them are backpacks. We have built some machine learning facilities to understand what products really are and fit them into our own hierarchy. So that’s our big effort, understanding more about the products we have, rather than understanding more about the users, where we are actually much farther ahead.
That has a number of applications. It goes beyond just showing shoes instead of backpacks—it also allows us to present different kinds of search interfaces. So, if you came along and said you were interested in Flight of the Conchords, the New Zealand musical comedy act, we could say “Do you want a T-shirt, a CD, a DVD?” Even better, we could predict, “You don’t seem to order a lot of DVDs, so I’m going to reorder the search results.” There’s a bunch of stuff we could do with that.
X: How is it that you think you’ve been able to stay successful even in such a terrible economy?
SA: I was at a presentation a few weeks ago and they were saying that even in a down economy, the “affordable luxuries”—a purse, a skirt, clothing, shoes—have grown market share year after year. Jewelry, autos, homes, and other very expensive things have plummeted. We happen to be very well positioned because a lot of the affordable luxuries are where people are still spending high amounts. Plus, we are relatively small, so the type of growth we can effect by just drawing in a few more Google visitors is enormous.
X: I have to confess that I’m not a huge user of StyleFeeder, because I’m not usually looking for product recommendations. I know what I want, I go in to a store or a website and buy it, and I’m out.
PJ: That’s because you’re a man. And it’s actually been frustrating over the years, as we were raising money, trying to talk to venture capitalists—who are almost all men and who shop the same way you do. Women browse a lot more and tend to be less direct, more circuitous. There is this idea called the “savannah hypothesis of shopping” that says men are hunters and women are gatherers, and we totally see that on the site. As you rate products on the site, and as you add things to your style feed or wish list, those are strong signals, and we can predict your gender with an accuracy rate of 95 percent just from these things you do as a user.
SA: But even for you, if you knew exactly what you were looking for, like this exact Bell bicycle helmet, and the ones we show you aren’t quite right, right below it we are going to show you another 15 items, and there is a high likelihood that one of those will be something you like. So it’s adding value, even for you.
X: You make money mainly on the affiliate fees paid by online retailers, right?
S: Yes, 80 percent of our revenue is from affiliate fees. We are in the top 2 or 3 percent of publishers within networks like Commission Junction and LinkShare, which puts us on a par with sites that have been around much longer.
X: I’m not familiar with those names.
PJ: By and large we don’t have direct relationships with all of these thousands of merchants. Affiliate marketing is a whole industry that handles transaction tracking and management, data feed uploads, all of that. Commission Junction is a huge player, as is LinkShare.
SA: If you tag your products with the right affiliate code, they will track the online reporting for clicks and revenue. A lot of the value these networks can add is in facilitating introductions. We started out with no relationships in the retail industry. Now, with their help, I’d say that at the top 70 or 100 retailers—a BlueFly or a Macy’s or a Target or an Amazon or a Wal-Mart—I know somebody at all of those companies, and they are giving us more favorable commission rates. We’re pushing close to 8 percent per transaction with some of them.
X: It sounds like you really couldn’t exist without the affiliate marketing industry.
PJ: StyleFeeder couldn’t exist without cloud computing, venture capital, or affiliate programs.
SA: Well, it could exist, but it would just require a different capital structure. You’d have to have more people like myself calling on all of the retailers and setting up the data streams one at a time.
PJ: And instead of $3.5 million in venture capital you’d need $15 million.
