This week, following a tumultuous 18-month courtship, Microsoft and Yahoo finally consummated a deal to meld their search services and online advertising technologies. According to Kosmix co-founder Venky Harinarayan, however, the only way to alter the online search game and gain market share is to “change the definition of search.” Unlike Google, which ranks sites according to popularity, Kosmix drills deep into particular topics, or “verticals,” drawing on content across the web to create what Harinarayan refers to as a “browsing experience” for the user. Among the products the company has launched are RightHealth — the number-two health information site on the web — and MeeHive, which allows fine-tuned customization of news content. Harinarayan spoke with India Knowledge at Wharton about where online search applications — and Kosmix — are headed.
An edited transcript of the conversation follows:
India Knowledge at Wharton: How would you characterize where Internet search engine competition is going, with respect to Google as the existing market leader, Bing, Cool, LeapFish, Yahoo and WolframAlpha? And how does Kosmix fit in?
Harinarayan: Looking at the broad picture, we have seen a couple of trends emerge over the last four or five years which are starting to create some amount of change in the web space…. If you look at the web 10 years ago, it was largely a collection of web pages. People wanted to go to web pages — that is what they wanted to do.
Over the last four or five years, we have started to see applications [that are] less about [static] web pages — [such as] Facebook and YouTube…. The web has really evolved. Search has not really evolved, however, over the last 10, 15 years; it is the exact same sort of product that it was 15 years ago.
India Knowledge at Wharton: So, the web has been transforming but search has not evolved over the last decade?
Harinarayan: Yes. Ten years ago, the consumer was largely anonymous, right? You would use a dial-up modem, find a web page, and you could not afford to have rich media or any of those things because connectivity was all narrow band. Now, with most people connecting with broadband, what you see is things like Facebook, YouTube, Flickr, and people can even upload a lot of things. People want to have a presence on the network; they want people to know they have an identity.
We have seen a complete shift in terms of how people use the web, but search has not changed at all in these last 10 years. You can say that Yahoo has gotten smarter … and so on, but the information-finding experience has really not changed — and that is, I think, the opportunity.
Search does what it does incredibly well today, and so it is hard to sit here and say, “Hey search is broken.” That is the reason I believe it is going to be hard for the Bings and the Microsofts and the Yahoos of the world really to start capturing market share against Google without changing the definition of search.
If you want to win, in our opinion, you have to change what search means. You have got to change it or at least do it much better so people say, “Hey, yeah this is something completely different.” But so long as you are playing with the search positioning of just finding, helping to find the better 10 pages — a better page, or best page on the web for your query — I think that is a really tough hill to climb at some level.
India Knowledge at Wharton: Now okay, so just to build on that first question, how does Kosmix fit into the long-term view of potentially changing what search means, versus other players who have been getting a lot of press, like WolframAlpha?
Harinarayan: Wolfram is definitely aligned with the way we look at the world in essentially trying to redefine what search means, even if it is a small piece of search. So, it is a completely different experience, but it is not like trying to give you a better 10 pages on the web. They are providing you an engine which effectively it is sort of an extension of Mathematica at some level, which is a computation engine…. It has nothing to do with the web, really; it is an engine, they have lots of facts, they try to churn through the facts, run some inferences and sort of give you an answer for your question.
India Knowledge at Wharton: Right.
Harinarayan: So, it is a different model than search entirely. In fact, I would say we should not even call it a search engine; it is sort of a computation engine –which is what they call it, right?
India Knowledge at Wharton: What about meta-searching versus search aggregation?
Harinarayan: Kosmix has taken meta-search to the next level. The old meta-search engines were all about taking three search engines — Google, Yahoo and Microsoft — and trying to give you the top 10. And I think fundamentally what we are seeing is that consumer positioning is done — searching Google, Yahoo and Microsoft, the way they are today, is a pretty good experience. It is a pretty mature market; the market shares are not going to change very much. It is going to be incredibly expensive to change those market shares. It is a conventional, mature market. Typically, market shares get set early in the market’s evolution. We are in the late stage of a market here. You do not go changing market share points willy-nilly — it takes a lot of dollars, and you can move your points up and down.
So, top 10 results — that is the game. It is never market share gain. You have got to have a huge amount of resources like Microsoft….
India Knowledge at Wharton: How does a company make a user switch in search without AdSense throwing US$100 million at it?
Harinarayan: What we are saying is, “Hey, if you want to find what the homepage is for FitLiving.com, do not come to us — that is not what we are good at. For that you should go to a search engine.” On the other hand, if you want to know more about Kobe Bryant, we give you a much better experience than you would get by typing “Kobe Bryant” into a search engine…. The bet we are making is long term: People are going to start separating these cases in their mind…. I think if that happens, we [will be] a very, very successful large company.
India Knowledge at Wharton: What is your source, in your mind, of sustainable competitive advantage?
Harinarayan: If you look at the positioning that we want to take, which we feel is being satisfied by search today but not really well, it is really the “browse, explore” positioning. Today, that positioning gets expressed in search. It is not a very good experience because what happens — let us say you are looking for Kobe Bryant, if you want to just know more about this guy? So you go and then type his name into a search engine and what you see is the first result for Kobe Bryant, wanton soup or whatever topic you want to know more about, and fundamentally you are restricted to the sites that came up first on the Google results.
That site is largely human created, so fundamentally your experience is what one person on the other side thought was the right experience. And our bet is basically by using the entire web, we can actually give you a much better experience than, let us say, what one particular site can give you.
Fundamentally, the vision here is when you want to browse, it is very hard to browse with a search engine, and much easier to browse with an experience like ours, so that is the underlying vision which we take forward.… The model we have taken is to say, “How do we enable browsing no matter where it happens on the web.” And that is the underlying core thinking behind what we are doing.
Looking at the web today, we see three areas in which browsing happens. One is, like I said, you go to a search engine and you type in something and you go and explore. So you have strong intent, you really to want to browse wanton soup. You want to know how to make it, blah, blah, blah — you just go into a search engine and you type it in there.
The second place you browse is when you are actually reading a content page. Let us say you are reading an article about how Barack Obama made an address to the American Medical Association today. You want to know a little bit more; while in context, how do we show you the right set of links to make those connections? That is not being done well today.
And the last [situation] is when you have a strong interest in a topic — for example, you are very interested in technology, so every morning you get up and you check [certain sites]. That is, again, a place where you start browsing.
That is fundamentally what we want to enable — and when you look at the set of products that we have in our portfolio, those are the three sort of pillars along which we think about our business. It is intent based, which starts with search. It is context based, which is in a content page. And its interest based, which is where MeeHive comes into play.
India Knowledge at Wharton: Given your underlying fundamentals, can you explain a bit more about RightHealth and MeeHive?
Harinarayan: Let me give you some history. We started with RightHealth as our first sort of property where we were embracing this notion of “browse.” And health is a very important area for browse, because there is so much information along different facets of any topic — for example, if you are interested in diabetes, you can see that there is some very basic information there for someone who has no idea what the symptoms are and what the conditions are.
For a chronic sufferer of diabetes, you start getting into a lot more deep content like you could see in many medical journals — you can have alternative medicine treatments, you can have what your doctor will be reading, which is again very, very interesting. So fundamentally, when you have so many different facets, a part of the challenge, a part of the way you enable browse, is by just presenting all these facets in a very engaging fashion. And in health, especially, it becomes very important, because people like to drill deep but they like to do it in a very guided fashion.
And so we launched RightHealth. We got very good, positive feedback. It is sort of our first vertical, and we have done really well with RightHealth. It is the number two site on the web today for health information. We do around six to seven million unique visitors a month. So it has been a really, really good outcome for us.
Once we sort of proved the model with RightHealth, we decided to go horizontal and apply this technology and the product to any topic, whether it be diabetes or whether it be wanton soup. And it was a much harder problem than we thought, because once you pick a vertical, you make some choices, you make some divisions of the architecture that even without you knowing it are fundamentally vertical-limited. And so we had to change the entire model; it took us around six to nine months to re-architect the system to be able to do something truly horizontal. We launched the product last December, the kosmix.com product, and that is how we have evolved.
The horizontal product also makes it easy for us to build more verticalized experiences on top, and so you will see us do some of that but it will be in context of Kosmix. For example, the finance part of Kosmix might start getting really, really rich. And we will start creating homepages for different verticals like finance, sports, and things like that.
India Knowledge at Wharton: Like what Yahoo and Google do already?
Harinarayan: Exactly, but we do it in a bit of a fairly automated fashion.
India Knowledge at Wharton: Beyond ad sales, are there any other revenue streams that you are entertaining?
Harinarayan: At this point, it is not in our best interest to be charging consumers for anything. So, it will be primarily an ad-based model. We will definitely work with different models of advertising. I am happy to get into more detail if that something that you would like to discuss.
India Knowledge at Wharton: A brief overview of the different models, if you are able to discuss them, would be interesting.
Harinarayan: Sure. I think we have got three models that we really focus on; one is the display advertising which we use in health, especially. We have paid listings which we partner with Google and a couple of others on. We also use our own modules, called sponsored modules. So everything you see in Kosmix on the page, on the topic page, is a widget. And we also let advertisers sponsor widgets — and we make it very clear to the user that it is a sponsored widget.
India Knowledge at Wharton: Is the plan to get more Kosmix users currently word-of-mouth or are there other things [planned]?
Harinarayan: Absolutely we do a combination. So you have got word-of-mouth, you have got — we actually get our pages indexed on the search engines which also works well for us. And like I said, we have started to connect to the contextual sites, or getting our pages linked to from other content pages, as a driver of traffic.
India Knowledge at Wharton: And in terms of the unique visitors that you have, is there a breakdown,? Are you seeing them more in the U.S. or perhaps other countries?
Harinarayan: Actually, it is largely U.S. We get a fair amount of traffic from India, which is maybe the only other surprising or non-surprising piece. But I would say the U.S. is still the number one right now. And we get also a fair amount of traffic from Europe at the end of the day. But I think India sort of spikes more than you would expect.
India Knowledge at Wharton: I see. And then if you could chat a little bit, about going from vertical to horizontal and how MeeHive fits in all of this?
Harinarayan: Absolutely. Like I said, the fundamental vision here is browse. We believe the notion of topics is central to the browse sort of model…. MeeHive is the way for you to browse your favorite topics, keep track of your favorite topics, topics that you are interested in for a long period of time. How do we then give you the experience around how you browse these topics?
That is a very central piece of our thinking around browse, and you will start to see us integrate MeeHive, Kosmix into one sort of unified experience, just like Yahoo has My Yahoo and the Yahoo.com sort of search. We will start to put these things together. For now, just to get product speed, it makes a lot of sense for us to keep them separate and execute against the use of the consumer value proposition as a standalone property.
India Knowledge at Wharton: Here is a good time to transition away from the technology [issues] and talk more about your experience as an entrepreneur. So I will start off by asking what have you, [Kosmix co-founder] Anand [Rajaraman], and your team faced as the biggest entrepreneurial challenge so far with Kosmix? How did you overcome it, and then what did you learn from it?
Harinarayan: The biggest challenge is getting people to take a long view and be patient. It is easy to say but very hard to execute. Sometimes it takes time to build a big, great company. You have got to be patient, you have got to believe in the long-term vision, you have got to believe in what you were passionate about…. Keeping people focused, keeping people excited, I would say is the largest challenge.
This is true for any startup. It always takes time to build great companies, and a lot of times you have to make some tradeoffs where you say, listen, I know that I can do this and that it might give me some short-term gains. But I really need to think about how I build this [into] a large company in the fullness of time. So there are many times in the short term and the long term where you have to make those sorts of tradeoffs, biasing it towards the long term.
India Knowledge at Wharton: When you say it is tough to convince people to take a long view and be patient, you mean not only your partners, but also investors and also employees you try to recruit?
Harinarayan: Absolutely! So a very simple example of this is what we call consumer experience. A lot of times it is easy to say, hey, let me just get some advertising dollars out; I can make a fair amount of money in the short term. And many times doing that actually conflicts with the consumer experience. So to sit up and say, no, let us do the right thing by the consumer, let us really make sure that the consumer experience is not sacrificed, is a hard decision. That is one example where it is counterintuitive but it is the right thing to do.
India Knowledge at Wharton: How did you and Anand go about finding the right talent at Kosmix, given what you just said?
Harinarayan: Talent sort of always goes in cycles. The way I look at it in the lifestyle of a company, going from let us say three [employees] to 10 is the hardest time. It’s just incredibly hard, no matter who you are, to convince your fourth employee to join you. It’s like pulling teeth, okay, because no one wants to join a three-person company and they do not have any idea what to expect, so it is brutally hard.
After you get 10 [employees], things start getting a bit easier. Basically, what happens is you have hopefully hired nine good guys; also your bar for the first 10 guys is incredibly high, because it is crucial that your early team is absolutely A plus. So, some of the toughest times we had was really hiring the first 10. Once you have crossed 10, as the magic number, what you see is that if you have done a really good job of hiring nine or 10 really good people, it starts getting much easier to convince the 11th to 20th to 30th to come in and join you.
Then you end up with a lot of problems when you hit about 40 or so. You start losing people or you start being unable to hire people who want to join smaller stage opportunities. None of them wants to join a three-person company; everybody wants to join in the 20 sweet spot range.
When you start hitting 40, you have to plot the horizons in terms of who you want. It also is the right thing because a lot of the folks who were really early stage junkies, they may not be the right folks for you [later on]. In my experience, you hit these [challenges] in every startup’s evolution but I would say by far the first 10 [employees are] the most crucial. You set the culture and you set the tone, you set all of that for the first 10.
India Knowledge at Wharton: What lessons did you draw from your experience with Junglee that you imported to Kosmix? Maybe [these are] lessons that you knew about [or] ones that you have just found that you have imported without knowing.
Harinarayan: I tell people that having had a dialup company before is a two-edged sword. It is really good in one way in the sense that you are much more efficient, you do not make the same mistake twice, you sort of know the lay of the land, so you can avoid the pitfalls. You are really much more efficient as an entrepreneur. You are very focused; you do not [get diverted] which you do as a first-time entrepreneur and so on.
I tell people that is actually a two-edged sword. It sounds like it is all positive but sometimes you need to be stupid, you need to be foolish, you need to do things that are not smart because who knows where the pot of gold is, right? So it is something that I challenge myself with all the time, which is — it is okay to do these things, even if I know that it maybe is not the right thing, every once in a while because you have to challenge yourself in that way. I think that is the challenge for a repeat entrepreneur.
It helps, you know, to be naive and to be optimistic about things even if they are wrong every so often.
India Knowledge at Wharton: Understood. Now I would like to know a little bit about how you and Anand first met? How the two of you, as you met, began to establish trust with each other …
Harinarayan: We met a long time ago. We met when we were in Stanford; we were grad students in the same dorm, Escondido Village. I think we met each other socially before we did professionally, in the same department. He was one apartment away from us, so we just got talking. He did not have a car, so I had to take him grocery shopping at that point. That is how we met.
We had the same advisor, so that helped a lot. We wrote a few technical papers together and that is, I would say, where the trust got built. We collaborated at Stanford with each other and started Junglee together, and it has been easy to keep the trust level and the relationship going. It is not the case that we agree on everything.
In fact, people here [think] we do not yet agree on much, but there is a strong degree of trust. If Anand is speaking about something, really passionate about something, I trust him 100%. Even if I do not necessarily fully agree with him, I almost say, hey I do not even want to think about it, if you are thinking about it, because I know you will do the right thing at the end of the day. And so I think it is the same with him as well.
India Knowledge at Wharton: I see. Are there any specifics you could offer about your best experience with Kosmix so far and your least favorite experience with Kosmix so far? Something where you felt really, really high and something where you felt not so great?
Harinarayan: Let us see. I would say that RightHealth doing as well as it has, has definitely been a really, really good experience at the end of the day. It has been a high but the good news is that the high has been ongoing. That is something I am really excited about.
The other thing which really excited us was when we first launched our product in December and we had a really nice story in The New York Times, thanks to Jodi. The feedback we got on Twitter and the Blogosphere was absolutely, incredibly positive. That was a real high for the whole team and for me in particular.
In terms of lows, it is hard to sit back here, because … you try to get the lows out of your system very quickly. Otherwise you cannot really do a company. Even [regarding] the highs, I try not to focus on the past because you just have to be thinking two steps ahead. I would say there were times, especially with MeeHive in the early days where we were not clear what we were doing, what we were trying to get done. There was a strong belief in the end point and it took us a little while to be able to articulate it, but I think I am really, really happy that we stayed the course.
In fact, yesterday I saw a tweet on MeeHive, and it was something like. ‘This is the best website I have ever seen in my life.’ It is not only the straight path which gets you to this best website. We have made a bunch of twists and turns along the road but I am really happy that we are here at this point. Some of those points when we made the twists and turns were questions you asked [about] why I am doing this. I am really happy that we persisted and actually are on the way to what I feel is a really, really great consumer experience.
India Knowledge at Wharton: I would like to get a little more abstract about being an entrepreneur, and ask generally how your roots as an Indian person have played into your experience — coming to Stanford, meeting Anand, starting Junglee and now doing Kosmix. When you sit back and reflect on that, how have your roots played in that movie?
Harinarayan: Actually, my dad was an entrepreneur. So I think that had a lot to do with it. That said, being an entrepreneur in India when he was, was 100 times harder than it is today in the U.S. You played with your own money; if you lost, you lost everything. And it was hard; it was in a socialist sort of environment, so you could not hire and fire. The mid-70s was not an easy [time] to be an entrepreneur. I have no idea why my dad did it, and my mom sort of made me promise I would never do that ever in my life. That is where we started.
But once we came here, the friction to starting a company is so low; it is something that at the end of the day is so easy to do. And also, I went to school at Stanford. I do not think I particularly went in there with the idea of starting a company. So, it is sort of an interesting story how we started Junglee. We came up with this idea that was really a technology idea when we started: How do you take all the sites on the web and integrate them in a very easy fashion?
The database group at Stanford, where we were, actually was doing something related, except they wanted to integrate 10 to 20 databases. So we went up to them and said, ‘Hey, guys, yeah, 10 to 20 is really interesting, but there are thousands, if not millions, of sources on the web. Why do we not focus on integrating them?’ This is just when the web was happening.
So we actually went and pitched to the database group: Shall we do a project — a research project on this? And they shot us down for a very good reason — because they had a lot of money associated with the old project. So we went into hiatus at that point. The idea went on hold.
Then that fall, there was a conference at Stanford and Yahoo Founder David Filo was out there. So I just went up to him and said, ‘Hey, we were noodling this idea. What do you guys think?’ And he said, ‘I love this. Why don’t you guys come over and talk to me?’ So we went over and talked to him. He wanted us to work at Yahoo and we said maybe we should just try this thing on our own. That is how we really started Junglee.
The story there was that at the end of the day, it did not start as an entrepreneurial thing; it started as something that we felt there was a need for that we would really like to do. We were the biggest consumers for this thing at the end of the day, if it happened, because there were all these great data sources, whether it be a retailer or whether it be a jobs database, that were simply not hooked together. So if you wanted to find something, you had to go into each of these databases and actually search for information. And that was really painful and hard to do.
When you start with a need and you feel like there is a real sort of opportunity around that need, that is the basis for how companies get created. Now, our path was like every entrepreneur’s path; it was not a straight line. It actually twisted and turned a little bit before we decided to do start something.
India Knowledge at Wharton: Coming to the U.S. and relocating in the bay area — was that kind of a unique place for you and Anand to do something like this? Were you able to draw on the diaspora in any interesting [way]?
Harinarayan: There are a couple things here. [Had I not been in the bay area,] would I have started a company? I do not know. It is just absolutely hard to look back and say. The truth is that the bay area made it really, really easy to actually start something.
Once we started something, it was incredibly helpful to be in the bay area, be able to draw on the other entrepreneurs there, a lot of them. [And] it was easier to get hooked into some of the folks in the Indian diaspora in the early days. They were much more open to working with us and it was incredibly helpful. That is it. I would just say that if you are [in] the bay area, everybody is open to talking to you. When we started Junglee, we went and talked to a firm called Interval Research. This was Paul Allen’s think tank.
These guys were amazingly helpful and they wanted nothing for it. They had so many connections. They connected us to our first board member and so on. People just like to help entrepreneurs in the bay area. The India diaspora [helped a lot also]…. It was obviously something that we leveraged a lot and it was extremely helpful to us in getting the company off the ground.
India Knowledge at Wharton: So just to probe on that Indian diaspora comment. Are there any stories you can recount or things you remember about drawing specifically from that diaspora that helped you in a specific way?
Harinarayan: Oh, absolutely. Initially when we started, the seed investor in our company was a Japanese gentleman…. We just cornered him [at a conference] because one of our founders knew him well…. We pitched him the idea and he ended up investing at the end of the conference, which was sort of funny…. We took him over to Stanford and showed him everything. He was very helpful in the early days.
We had made a whole bunch of connections into the Indian diaspora for sure. You know, a lot of the early employees we hired were either from Stanford or were people we knew from before, so it helped a lot.
India Knowledge at Wharton: Okay. What advice would you give to somebody who is currently in a PhD program — in your PhD program right now at Stanford, or somebody who is coming to the bay area and looking to roll up their sleeves, maybe not a technologist but still an entrepreneur? What advice would you give each of those people, based on your experience being Indian, now living in the U.S., having these two ventures? What would you tell them?
Harinarayan: I would say a couple of things. It always starts with a dream. That is what I tell people. If you do not have the dream, do not do it, because it just takes far too much out of you and it is not worth it…. So you want to have a really, really big dream in terms of the passion that you are willing to bring to it. The only reason to do it is because you cannot live with yourself unless you give it a shot. That is important because I have seen too many people start companies for the sake of starting companies, and that is the easiest way to fail. So, that is the number one thing.
It is not like it is a logical analysis. You just say, hey this is what I need to do, this is my mission in life, let me just make it happen, let me just do whatever the heck it takes. Everything changes when you start with a dream that you are passionate about and you really feel like you want to change the world. That is my [best] advice to people: Do not start a company for the sake of starting a company. Do something because you want to change the world.