hunch

A Conversation With Hunch Cofounder Caterina Fake

Next Story

Qik Indeed. Service Comes Built-In To The Sexy HTC EVO 4G Android Phone

New York based Hunch is on a bit of a roll. Users have now answered 50 millions questions on the service, which uses your answers to help give you recommendations for just about anything. Investors have taken note of their progress, and they recently closed a $12 million venture round from Khosla Ventures and previous investors.

I sat down with cofounder Caterina Fake last week to talk about Hunch and how it can help people make decisions about things. As people use the service over time Hunch is able to build out a very deep taste graph of what they like and dislike. It’s a sort of DNA of people’s taste preferences and a parallel to the social graph that they are building out on Facebook and elsewhere.

The lights are too bright and the sound echoes (we’re still working on our new studio), but Caterina gives deep insight into what Hunch is all about.

And we also talk about Yahoo. Caterina was an executive at Yahoo from 2005 – 2008 after her previous startup, Flickr, was acquired. Yahoo has an uphill battle, she says, and we spoke about their need for a product visionary:

The full transcript is below:

Mr. MICHAEL ARRINGTON: I’m here with Caterina Fake, the co-founder of Hunch. Hello.

Ms. CATERINA FAKE (Co-Founder, Hunch): Hello.

Mr. ARRINGTON: Thanks for coming in.

Ms. FAKE: Thank you.

Mr. ARRINGTON: You guys just raised a big round of financing a week ago from Khosla Ventures.

Ms. FAKE: Twelve million dollars.

Mr. ARRINGTON: Yeah and we covered that, but I thought it was probably time to maybe talk to you a little bit more about Hunch and do a deeper dive than we have in the past with the company.

Ms. FAKE: Yes.

Mr. ARRINGTON: So, OK, first thing is, what is Hunch and why does it matter?

Ms. FAKE: So, the way Hunch works is Hunch learns about you. Hunch asks a bunch of questions which we call Teach Hunch about You. Those are questions about demographics, beliefs and values, your aesthetics, your political views, all those kinds of things and it’s meant to be fun. There’s a kind of a little module up in the corner of the front page if you go in and, you know, it’s everything from, do you live in the country or the city or the suburbs. Do you cut your sandwich down the middle or diagonal? You have you ever used a fake ID when you were underage? That kind of thing…

Mr. ARRINGTON: And that tells you Hunch…

Ms. FAKE: And it tells you all of these – all of these like really interesting things and it enables…

Mr. ARRINGTON: If I cut my sandwich down the middle or diagonally, let’s say kind of diagonally, what does that tell you?

Ms. FAKE: Yeah, there is this tremendous database that we have of all of these into like fascinating correlations and it turns out there’s certain things that are like the example that I just gave; if you have used a fake ID when you were an underage you’re significantly more likely to be an entrepreneur. It means that, you’re like kind of like more…

Mr. ARRINGTON: Does everyone use a fake ID? I use plenty(ph). Could you use…

Ms. FAKE: Well, like the Silicon Valley Entrepreneurs? And I used…

Mr. ARRINGTON: You used one?

Ms. FAKE: I used a fake ID when I was underage.

Mr. ARRINGTON: Is there a (unintelligible) in Canada?

Ms. FAKE: There – I’m not Canadian. I’m actually American.

Mr. ARRINGTON: Oh you just lived in Vancouver when you guys started?

Ms. FAKE: Yeah. Yeah.

Mr. ARRINGTON: OK

Ms. FAKE: Yeah. So, you can learn all these really interesting things and, Hunch really is a way for us to give you the best recommendations possible. It’s a bit like Amazon or Netflix recommendations for everything and…

Mr. ARRINGTON: What is – that’s for? Movies? Restaurants? Vacations?

Ms. FAKE: It – I wouldn’t actually – no, I wouldn’t actually limit it to anything in particular. I mean, it should work for everything. It should work for, I need a hotel in Dallas. I need to find a philosopher. I need the New York Time’s bestseller to bring on vacation with me, we know everything. It should work for those things that especially are very taste-oriented and I think that the thing that really distinguishes Hunch from other efforts out there, there’s a lot of kind of products that are, attacking a kind of like a similar problem is that we are trying to create this taste graph for the whole web. We’re trying to learn the tastes of all of the people out there and…

Mr. ARRINGTON: That’s obviously a planned social graph, right?

Ms. FAKE: Which is a planned social graph, right? It’s analogous like the taste graph we’re saying is something that’s analogous to like the social graph. But from our perspective, the social graph is actually less informative and actually gives you less valuable information on you than what we’re calling the taste graph because you may – I may be in contact with my co-workers, who are kind of like male engineer types and, with my mom, I have a very close relationship with. But our tastes are very different, the things that we like, the sushi restaurants or (unintelligible) that we’d be interested in…

Mr. ARRINGTON: Yeah.

Ms. FAKE: Or the clothes that we would wear and so, what we’re – our assumption is that there’s people out there who share similar taste. They have a similar aesthetic to you or they have, say, you’re kind of looking for a blog or a news show, your political position or political stance would inform that choice as well. So, that’s really what…

Mr. ARRINGTON: And this actually works?

Ms. FAKE: And this actually works, yes.

Mr. ARRINGTON: You found statistically relevant results…

Ms. FAKE: Yes. And it’s a completely data-driven kind of like algorithmic side and so the way that Hunch evolved was we launched last – June of last year and I’d say that this first – the first part of this phase of the product’s development, we had a kind of a chicken and egg problem in order to make really good recommendations, in order to understand the taste graph. We had to gather all of these data. We had to have information. We had to get kind of a cold start problem. And so we’ve spent the past – how long has it been – seven or eight months now gathering information about people, learning about people and we have 1.5 million uniques(ph)…

Mr. ARRINGTON: Yup.

Ms. FAKE: In the past couple of months, each month and…

Mr. ARRINGTON: So, 1.5 million people a month are coming to the site…

Ms. FAKE: Are coming to the site and…

Mr. ARRINGTON: And answering these questions…

Ms. FAKE: What we have – I mean, we do have 200,000 – I would say the best data that we gather from the 200,000 registered users who were – have been answering an average of 140 of these questions. So, we have over 50 million questions answered in our site.

Mr. ARRINGTON: Well…

Ms. FAKE: So, it’s an incredible amount of data. And now, I’d say we’re in the position where we can actually use this data. We can actually make assumptions. We can make – we can draw connections between people and products and services and all of those kinds of things.

Mr. ARRINGTON: How do you – well, actually, let’s take one step back and then talk like the company was different before you started because you’re a co-founder, and you actually joined little later and what…

Ms. FAKE: Yeah.

Mr. ARRINGTON: How – what – did you…

Ms. FAKE: That’s right.

Mr. ARRINGTON: Move the direction and what was it before and why did it change?

Ms. FAKE: So what happened with – what happened with Hunch is it was – it was originally founded by Chris Dixon and Tom Pinckney and Matt Gattis. And they had come out of – their prior company was SiteAdvisor which they built in, I believe, 2005, 2006.

Mr. ARRINGTON: Yeah.

Ms. FAKE: And they were – they sold it to McAfee for $74 million. What they were doing was like kind of like security – a security site which analyzed, you can kind of guess by the name, it’s like SiteAdvisor tells you the kind of the relative safety of visiting various sites. And so their initial product, the thing that they were developing originally was it was a decision tree model. It was kind of based on expert systems where a series of questions were asked and you got an answer at the end. And so they started off – they kind of hoping to fund experiments that they try that did not succeed including, sort of like a thing called pure printer fixer(ph) or you had to go down this decision tree, like is it plugged in, it’s the toner in the cartridge. You know, they basically kind of take you down this path and observe a diagnostic tool. But they knew that there is something very interesting about destruction in this model learning things. And then there is this – there’s this other part of it where they, ask questions, some of which were amusing, some of which were kind of very serious. Those were the taste-oriented…

Mr. ARRINGTON: Yeah.

Ms. FAKE: Questions that I would say. And at some point, they had done – they have been doing this in an entirely automatic way.

Mr. ARRINGTON: Yeah.

Ms. FAKE: They have kind of like – for example they had sucked in the entire, products database of Amazon, for example. And it felt computational, it felt mechanical. It felt as if a computer had done it. And at some point they realized that the only way that this was going to work, that the only way the Hunch was going to work, was as if it were a user-generated content site and that there had to be human input.

Mr. ARRINGTON: Yeah.

Ms. FAKE: There had to be people contributing to it. And that was coincidentally the week that they figured that out. I was introduced to them by their venture partner, one of their investors is Bessemer. And so I kind of walked in just at the moment and they realized that this had to be man plus machine. It had to be human-added content as well as sort of very, (unintelligible) kind of computational algorithmic stuff on the back end. And so that of course, is my background and that’s kind of one of the strengths of my, part of development so…

Mr. ARRINGTON: All right. So looking forward, how do you know it’s working? Like what’s the feedback from the users that tells you…

Ms. FAKE: Yeah. I mean, we do…

Mr. ARRINGTON: We’re helping them.

Ms. FAKE: We we measure kind of internally as our success rate. So we give a recommendation and the user will say do you like this, yes or no. Well, no if they click through a to read more about it, if they’ll, go actually and go purchase the product. Those were – those are the strongest signals that we have. And, we were starting on – this is our so-called success rate has been increasing over time. When we launched it was, in the 70′s. You know, like 72 or 73 percent success rate.

Mr. ARRINGTON: And that’s based on the user does something with the recommendation.

Ms. FAKE: Yes, yes, like in a positive direction I guess. That means a positive kind of – in a way that’s increased of, you know, until like the 90′s now. And so we have a good 90 percent success rate. And so that’s how we measure our success so…

Mr. ARRINGTON: It’s interesting that you allow people – it’s so – your recommendations are based on the data that you get…

Ms. FAKE: Yeah.

Mr. ARRINGTON: From people on building the – well, you’re calling the taste graph.

Ms. FAKE: Yeah.

Mr. ARRINGTON: But you allow anonymous users to go in and use the site without being logged in.

Ms. FAKE: Yeah.

Mr. ARRINGTON: And do you just give them that pretty good answers and why…

Ms. FAKE: Yeah. I mean, there’s a significant increase in our success rate obviously when people who log in when they answer questions when we get to know them over time. And I do think that, if we were to re-launch today one of the things we would very strongly consider is that it only be a logged in experience. I mean, I do think that when we launched it’s one of those things that you learn, and we’re – it’s a very new product we are inventing in as it kind of goes along and, this got – this to me, personally is of – is, kind of the most gratifying product to build as the one where you’re not really sure or like how to build and it’s not really clear how it’s going to turn out. You do a lot of experimentation. And I do think that one of the things that we would have considered from the outside is making it just logged in. I mean, I do think that, it very clearly improves to an incredible degree with our experience…

Mr. ARRINGTON: So why not change it now or just…

Ms. FAKE: Yeah, yeah. That’s one of the things that’s sort of under consideration.

Mr. ARRINGTON: OK.

Ms. FAKE: Well, we do have an amazing increase in the number of users that were coming in through SEO. And so it’s a trade-off. Because you are able to give people pretty good answers, pretty good recommendations without them having answered 140 questions as the average walked-in user has. But it’s not nearly as good.

Mr. ARRINGTON: Something I wanted to ask you. You were the co-founder of Flickr…

Ms. FAKE: Yeah.

Mr. ARRINGTON: And – and we know how that turned out. And so many people love Flickr and you’re a bit of a celebrity founder now. You definitely entered that.

Ms. FAKE: Yeah.

Mr. ARRINGTON: Is it easier to grow a startup as a celebrity founder or harder…

Ms. FAKE: You know, that’s an interesting question…

Mr. ARRINGTON: Or is it – the more tension there’s more expectations or…

Ms. FAKE: There’s more – there’s a lot of – I do – I do think that’s the case. There’s definitely more tension which is…

Mr. ARRINGTON: Yeah.

Ms. FAKE: Which can be good and bad. I do think that a lot – one of the things that you have to be on guard against is not taking risks because I think that you can, if you’ve…

Mr. ARRINGTON: Yeah.

Ms. FAKE: If you’ve had a big success previously, you can stop taking risks. You can…

Mr. ARRINGTON: In fact, psychologically, you’re…

Ms. FAKE: You’re kind of guard it against – you want to kind of – you keep what you’ve got, right?

Mr. ARRINGTON: Yeah.

Ms. FAKE: You want it, and so it makes you less willing to take risks.

Mr. ARRINGTON: So how you do fight that?

Ms. FAKE: I mean, you just have to very deliberately do that. And I think in a lot of ways, you have to really kind of like focus in and kind of filter out the noise and all of the kind of attention that you could potentially get drunk on.

Mr. ARRINGTON: Yeah.

Ms. FAKE: Like people you know, if you kind of – it happens to – it happens to a lot of successful companies, Not just individuals or founders but companies themselves if you start believing in press releases. It’s a very dangerous situation that you want to get out of. And to me, I think it would’ve been – I’ve seen many – many second time, third time entrepreneur founders going down the same route that they had gone down before without a lot of variation.

Mr. ARRINGTON: Yeah.

Ms. FAKE: And that was something that I’m very much wanted to avoid and Hunch is obviously vastly different from Flickr. And I think part of that was informed by the time that I spent at Yahoo. When I was at Yahoo, the most interesting thing that was going on there was social search. At that time since I was there, 2005 through 2008, the most…

Mr. ARRINGTON: And you weren’t running Flickr. I mean, that was handed off, right? I mean, you were…

Ms. FAKE: Yes. What happened was that was when we came in, we came in – we made sure that Flickr which has a very good outcome, that Flickr stayed Flickr. We wanted to make sure that it did.

Mr. ARRINGTON: Yeah.

Ms. FAKE: It’s very common for a company to be acquired and to lose its heart and soul and to kind of get – get lost. And so, you know, we spent a lot of time protecting it, making sure that…

Mr. ARRINGTON: Yeah.

Ms. FAKE: You know, the heart of Flickr remained. And then once that, once I felt as if Flickr was on the right path, then I branched out. And the most interesting thing going on, as I was staying, was social search and, at that time, and this is no longer the case, unfortunately, but Yahoo is going up against Google in search. And so as a result – and they knew what their strong suits were.

Mr. ARRINGTON: Yeah.

Ms. FAKE: Yahoo knew that a strong suit was a social. They had logged in users. They knew a lot about their users. Their users were people that came back. They, used a lot of different products an average of like two and a half products, like whatever Yahoo Finance or, mail or IM or what-have-you. And you know, and as a result there is an incredible amount – amount of social information that could be used in search. And so that was a lot of the work that I was doing while I was there, I was working in social search. And then subsequent to that, I worked on the problem of actually developing product at Yahoo as well…

Mr. ARRINGTON: Yeah.

Ms. FAKE: Because they know where the problem was. You know, obviously coming from the startup going into a large organization like Yahoo is that often the obstacles there’s no innovation deficit in a company like Yahoo. There is an incredible amount of invention and creation, but often what happens is the organization gets in the way.

Mr. ARRINGTON: Yeah.

Ms. FAKE: And so then, subsequent to kind of working in social search I worked on – I worked on, how do you get products built.

Mr. ARRINGTON: Yeah.

Ms. FAKE: Within a big company like that so…

Mr. ARRINGTON: I’d love to talk to you a lot more about that, but since you brought up Yahoo I will go a little deeper into it. I thought I’m getting into too much of a sensitive area…

Ms. FAKE: Yeah.

Mr. ARRINGTON: But when Yahoo changed CEO’s at the end of ’08…

Ms. FAKE: Right.

Mr. ARRINGTON: And they hired Bartz, Carol Bartz. A lot us were rooting for Yahoo to hire somebody who was product-focused, true product visionary. Microsoft had Bill Gates, Apple has Steve Jobs.

Ms. FAKE: Yeah.

Mr. ARRINGTON: There are other examples as well although not many.

Ms. FAKE: Yeah. But I agree with it, by the way, yes.

Mr. ARRINGTON: You agree that maybe…

Ms. FAKE: Yes. The product you’re in.

Mr. ARRINGTON: Somebody – and they didn’t. And they hired somebody who accomplished – accomplished executive…

Ms. FAKE: Yeah.

Mr. ARRINGTON: But did – and then they subsequently have not hired a true product visionary that I can see even…

Ms. FAKE: Yeah.

Mr. ARRINGTON: Even under Carol. Do you agree with like – you sound like you’re very pleased…

Ms. FAKE: I agree, absolutely. And I do think that one of the things that they should have done exactly is bring in a product person. To some extent…

Mr. ARRINGTON: Weren’t some of those people already in the company?

Ms. FAKE: Interestingly, yes. I mean, some of those people were already in the company and interestingly, I actually…

Mr. ARRINGTON: But they’ve all left now, right?

Ms. FAKE: I actually think that Jerry Yang – Jerry Yang was one of the best product people.

Mr. ARRINGTON: Yeah.

Ms. FAKE: And I – during the time that I was at Yahoo, all of the best product decisions I think, that were made during the era…

Mr. ARRINGTON: Yeah.

Ms. FAKE: During the period when I was there were made by Jerry Yang and all of the things that he funded – he funded the Yahoo Developer Network and it was a great example. He funded Yahoo Answers. I mean, all of the successful products that immersed out of there and all of the great…

Mr. ARRINGTON: What do you mean by funded? That he green lighted it?

Ms. FAKE: He had to green light it. He has a discretion pedestal, thus a discretionary project where he can green light any project from the company…

Mr. ARRINGTON: OK.

Ms. FAKE: As a founder, you know, chief guy.

Mr. ARRINGTON: Yeah.

Ms. FAKE: And everything that he green lighted and funded were, very successful. So, in some ways it is interesting because during the brief, you interregnum, you know between Terry Semel and Carol Bartz, he was the CEO. But unfortunately during that time it wasn’t – the focus was not on building our products it was on fending off Steve Ballmer in Microsoft so…

Mr. ARRINGTON: Well, why can’t he be that person today? Is it – he’s damaged now? He’s…

Ms. FAKE: I don’t know, I don’t know. I mean, I don’t have enough view. I guess, good enough view into the workings of the company now under Carol Bartz right now.

Mr. ARRINGTON: Do you think Yahoo can – can, not will, but can re-emerge as…

Ms. FAKE: Re-emerge? It’ll be a long uphill battle.

Mr. ARRINGTON: Yup.

Ms. FAKE: I mean, it’s – it’s – it’s sad to me because I think that Yahoo is in a not a very strong position as it was even when I started there in 2005. And I think the industry’s wish(ph) for it because I think a strong Yahoo is really good for the industry as a whole.

Mr. ARRINGTON: Some of your – I’ll call them competitors, they’re not really product competitors but guys like Aardvark and others been taken out and then acquired.

Ms. FAKE: Yeah.

Mr. ARRINGTON: Have you guys had some acquisition pressure that you’ve…

Ms. FAKE: There has been – I mean, there have been kind of, you know, nibbles…

Mr. ARRINGTON: Yeah.

Ms. FAKE: You know, kind of around the edges but we are not really interested in, you know, doing in acquisitions.

Mr. ARRINGTON: You don’t want a $50 million exit?

Ms. FAKE: No.

Mr. ARRINGTON: You want what?

Ms. FAKE: Yeah. I mean, I guess – you know, I – both my co-founders…

Mr. ARRINGTON: Right.

Ms. FAKE: They sold SiteAdvisor to McAfee. I sold Flickr to Yahoo. You know, none of us are really interested in, you know, a kind of an acquisition. You know, we really want to have a company that we built and, you know, and swing for the fences and go ambitious. And I think that Hunch we picked a very good problem. I think that we picked a very ambitious area.

Mr. ARRINGTON: Yeah.

Ms. FAKE: And so I think that it the possibilities are tremendous for the company.

Mr. ARRINGTON: Do you think – I am speculating here but if – let’s say I am a regular Hunch user and – I mean, I use it twice a month, three or four times a month.

Ms. FAKE: Yeah.

Mr. ARRINGTON: What does this got? The 200,000 hardcore users, are they using a – you said they’ve answered 150 questions each.

Ms. FAKE: Yeah.

Mr. ARRINGTON: They are using it pretty regularly.

Ms. FAKE: Yeah, very regularly.

Mr. ARRINGTON: Once or twice a month it sounds like – yeah, OK.

Ms. FAKE: Very regularly, yeah, yeah, and more in.

Mr. ARRINGTON: After a year or so of using it you have my taste graph.

Ms. FAKE: Yeah.

Mr. ARRINGTON: You almost have my taste DNA.

Ms. FAKE: Yes.

Mr. ARRINGTON: Like you can almost describe me psychologically…

Ms. FAKE: Yes.

Mr. ARRINGTON: Demographically as a – you know, as a human being, right?

Ms. FAKE: Right. Yeah. And the other thing too is that…

Mr. ARRINGTON: Does that – that’s still scary.

Ms. FAKE: What – you know, one of the things that we are interested in is actually allowing you to take your taste profile elsewhere. So, you should be able to, for example, we are going to roll out some stuff where it will help you – it will help you pick out, you know, who to follow on Twitter. It will help you – you’ve…

Mr. ARRINGTON: Choose the movies that I want.

Ms. FAKE: Choose movies that you want to see. It will be able to – you know, imagine – imagine, you know, you could take your taste profile and you could apply it to eBay or Etsy. And you could take – you could take it…

Mr. ARRINGTON: Yeah.

Ms. FAKE: Kind of like anywhere with you.

Mr. ARRINGTON: I imagine advertisers will be eager to get their hands on the idea as well.

Ms. FAKE: Right, right. But we have no intention of actually giving it to anyone. It was – it would…

Mr. ARRINGTON: You want to sell it?

Ms. FAKE: It would be no. We wouldn’t be selling it…

Mr. ARRINGTON: But you – I mean, you will advertise in to this, right? I mean, that is the monetization model. It must be.

Ms. FAKE: The advertising?

Ms. ARRINGTON: Yeah.

Ms. FAKE: The monetization model is basically it is the same as – it is the same as sort of kind of intent gathering mechanism like Google for example is, obviously if you are looking for a camera and we help you find a camera and then you click on the link to go buy a camera…

Mr. ARRINGTON: That’s where you monetize…

Ms. FAKE: That’s – that’s where all the monetization…

Mr. ARRINGTON: But not – let’s say a studio coming to you and saying I want to market Avatar…

Ms. FAKE: Yeah, yeah. No way.

Mr. ARRINGTON: To all the people you know are going to love it. Why not?

Ms. FAKE: Yeah.

Mr. ARRINGTON: I actually you might enjoy being told about it.

Ms. FAKE: Yeah. Yeah. Well that would – you know, like to date, that has not been something that we’ve been saying…

Mr. ARRINGTON: I mean, that’s if – you know, you are (unintelligible) Blippy – I mean, Blippy it’s like…

Ms. FAKE: Sure, yeah.

Mr. ARRINGTON: Where you – to post everything you buy.

Ms. FAKE: Yeah.

Mr. ARRINGTON: That I mean – that they’re blunt about the fact they were going to sell that profile…

Ms. FAKE: Yeah, yeah. Now I know. I know – what…

Ms. ARRINGTON: All day. It does not – we are thinking…

Ms. FAKE: When I’m planning on doing that. I mean, I like Blippy. I mean, Blippy is a really interesting company and I think that in many ways they are doing something very similar to what Hunch is doing and that they are creating a taste profile.

Mr. ARRINGTON: Yeah.

Ms. FAKE: They are approaching it from a completely different…

Mr. ARRINGTON: Yeah.

Ms. FAKE: Direction and that they’re using your purchase history as a way of – you know, of basically creating a kind of taste profile.

Mr. ARRINGTON: Yeah.

Ms. FAKE: But I do think that this is going to be a kind of a trend. That the social graph, which everybody is very excited about right now will be in some ways replaced by the taste graph. And what Blippy is doing would be as also kind of creating a taste graph as well but from a different source.

Mr. ARRINGTON: Yeah.

Ms. FAKE: In your purchase history.

Mr. ARRINGTON: OK. So, no quick sale for you guys, that’s a promise.

Ms. FAKE: No quick sale.

Mr. ARRINGTON: We’re going to see how this plays out.

Ms. FAKE: No quick sale, yeah. That doesn’t seem like it’s in the cards.

Mr. ARRINGTON: Congratulations on the round. Thanks for taking the time with me.

Ms. FAKE: Yeah, of course. All right. Thank you.

blog comments powered by Disqus