Yesterday I wrote up part of my recent interview with Chris Dixon, an active seed-stage investor and the CEO of Hunch, a year-old consumer Web application company that Dixon formed with Flickr co-founder Caterina Fake.
Hunch is a giant recommendation engine that aims to help users make up their minds, be it about a camera or what favors to serve at a Palm Springs party. Right now, a registered user base of roughly 200,000 people contribute most of the questions; they also populate the answers, a la Wikipedia. (In fact, Wikipedia founder Jimmy Wales joined Hunch’s board of directors in December.)
Hunch also relies heavily on proprietary algorithms that gets smarter over time, so that ideally, when someone arrives at Hunch with a question in mind, they’re served answers by users with similar preferences to their own.
Dixon, 38, thinks Hunch can be as big a leader in search technology as Google and Amazon someday, once it gets over some initial kinks, like making it easier for people to find and understand the service. The second part of my interview with Dixon follows:
You’re connected to many angel investors. Why did Hunch go for venture capital straightaway, raising $2 million from General Catalyst Partners?
We actually raised money from GC and Bessemer [Venture Partners], because they were our investors in SiteAdvisor [Dixon’s first startup, sold to McAfee in 2006 for $74 million]. We just put the band back together because they’re great investors.
I understand that for Hunch to work, it needs to gather information about its users, but I was asked a lot of questions. I also found the site a little baffling when I played around with it earlier. Is that a concern for you or are you trying to attract a certain type of super-committed user?
No, we sort of need to simplify our message a bit. What we want to do is provide recommendations for everything, using algorithms and data for any kind of topic. One of the challenges we’ve had is that what we’re doing is new and different and people in the tech world want to put us in a category. Is it Yahoo Answers, or search, or shopping?
Caterina cofounded Flickr (sold to Yahoo in 2005) and it’s now become part of the Internet fabric but at first, people asked if it was another Shutterfly or Ofoto and why are these pictures public and why can’t you print them. I think we’re exotic seeming because we’re new, but we’ll also keep refining the service and hopefully explaining it better.
Can you do that for us now?
Sure. I think part of it, too, is the context. Your experience right now kind of depends on where you come in. If you come to the site and are just messing around, it can seem a little confusing. But if you come to the site, looking for a camera or a TV , or if you come in through a widget at Sears.com, it makes a lot more sense.
I didn’t realize you had widgets at the sites of e-retailers. Where else do these appear right now?
We have some [retailers] with widgets and a bunch in the pipeline. We have thousands out there at smaller retailers; the bigger ones take time. It’s like enterprise sales.
What we also have now is tons of data about people, like the correlations between what people say about themselves and what they like, and we’re working on interesting features that will let people get recommendations in different ways. There’s a lot in the pipeline.
You have 200,000 registered users but many more people visit the service without registering or contributing any information, right?
Yes, we had 1.2 million unique visitors globally last month.
Which is a meaningful uptick from the fall, when you reportedly had in the low hundreds of thousands of visitors. To what do you attribute that growth? The widgets? Better data?
The data has gotten better, the interface has gotten better. There’s more instant gratification. We’re also showing up more in Google results; that kind of stuff takes time because of how Google’s algorithm works.
Also, more and more people are embedding the widget out there. We’re just marching along. I mean, we’re trying to do something really ambitious, which is change how people get recommendations online. It’s a many-years-long project.
What’s ahead? Will we maybe see dedicated Hunch sites around topics?
No, we’d rather embed the functionality into someone else’s site. That’s our strategy. It takes years to build up Website recognition enough for people to remember to type your name into their browser. So for the next few years, the majority of people will encounter Hunch elsewhere first.
Including where else, content sites like about parenting or entertainment?
Right, like right now, we’re working with Entertainment Weekly, which has this “which vampire series would I like best” sort of thing. I think content sites are a natural fit.
And your revenue model is based on some sort of revenue split?
It’s exactly like Google’s business model. If you’re on Hunch and you’re looking for a camera and you like the camera and click on a sponsored link beside it, we get paid.
How much do you know about your registered users?
Well, we know by definition they’re people who are early adopters and like to contribute. A lot of them are also Wikipedia contributors. We also have people with particular areas of expertise, like people doing women’s fashion stuff, people doing obscure science textbook stuff.
How do you ensure that they know what they are talking about? Do you have any editors on staff?
We do have an editor who ensures that nothing pornographic or anything is up there, but the nature of user-generated content is that some of it is good and some bad. You just hope that another user will come along and edit out the bad.
And in terms of the machine learning, is there any easy way to explain how that works?
Sure, the simplest way to think about it is: a user ads a camera to the system. The machine learning part learns when to recommend that camera.
Who is on staff right now, mostly engineers?
We have 12 people, almost all of whom are computer science folks. I think eight of them went to MIT. Caterina does all of our product design. We also have one guy who we call the head of content and marketing and who spends a lot of time putting together these reports that we issue around all this cool data and that’s been very popular with bloggers, like how food preferences vary by political ideology. Turns out conservatives like iceberg lettuce and liberals prefer arugula.
Good God; I hope you didn’t send that to Fox News. Do you sell any of that data or plan to?
No, we don’t sell any marketing data. We think we have a much better business model than that. We’d like to think of ourselves as that “If you like this, you’ll like that” feature on Amazon. If you want to compare us to something, it’s that feature, taken to the extreme.