The following is the transcript of an interview of Dr. Grant Ryan, co-founder and Chief Scientist, of Eurekster, conducted in several sessions in November, 2006. Grant and Eurekster have a rich pedigree in social search, going back to GlobalBrain, a company that Grant spearheaded in the late 90’s. GlobalBrain was all about “Learning Search” – a technology that was ahead of its time – designed to help E-Commerce sites provide a better user experience by learning from users’ search and browsing behavior.
GlobalBrain was ultimately purchased by Snap.com, which morphed into NBCi, in April of 2000. After the dot com bust, Grant and his brother Shaun bought the technology back, and put it to use at SLI Systems (for “Search, Learn, and Improve”), another hot young search-related company.
In addition to his role at Eurekster, Grant sits on the board of SLI Systems, as well as several New Zealand technology agencies that raise capital and guide New Zealand’s burgeoning technology sector.
Eurekster has been offering customized vertical search engine solutions for just over a year, and Grant was kind enough to take an hour to bring us up to speed on the latest developments in their Swicki product line. Interviewing Grant are John Biundo and Eric Enge of Stone Temple Consulting (STC). You can also read our Grant Ryan interview analysis, where we discuss the implications of what we learned during our discussion below.
John: Grant we really appreciate you doing this interview. We have had a lot of fun learning about Eurekster as a company and your Swicki product in particular. I think that one of the things that would help us understand a little more is for you to give us a little background as to how you got into the search engine business in the first place.
Grant: My first foray into this was back in ’98 when we had a company called Global Brain that we ended up selling to Snap, which later became NBCi. So we kind of rode the whole dot com boom and then the dot com bust, and after the dot com bust, we bought the technology back and have been running the company that uses some of the technology that learns from users of e-commerce sites. That company is now running as SLI Systems.
I then started a social networking company before they were called social networking, at least before LinkedIn or Friendster or any of those guys, helping people find jobs through social networks down here in Australia and New Zealand and that’s still going and it’s profitable. And then by observing those two and what was happening with social networks and search, the idea for Eurekster really went off as a bit of a eureka moment, I guess.
The idea that communities can help each other find things and share information – that’s really the idea behind Eurekster, the combination of finding things and the social aspect. The first attempt was a deal with Friendster, which used social networks to hone in on what was relevant. And then we found out what was interesting was not actually anything that had to do with friends and friends of friends, but what was happening at the group level, which is how we came up with Swickis.
A Swicki is what we think of as a combination of a search engine and a wiki, or wikipedia-type things. So any one can go and create their own. It’s an extension of personal publishing on the web — in the same way anyone can create a web page or blog, now anyone can create a search engine. It’s democratizing what is a really an important, powerful part of the internet today.
John: Let’s talk for a minute about the value proposition of Swickis. Let’s look at this from the site owner’s perspective. Say you are talking to a site owner or a web master who doesn’t quite get the value. The question is why should I build a Swicki? What is it going to do for me as an asset on my web site?
Grant: The same question could have been asked way back when as to why any one would build a web site or a blog or do a podcast. Basically it is another form of relevant content that you help manage that is useful for your users. So people come to your blog because you have good information on stamp collecting or whatever it happens to be. You can also give them a search engine that has got a stamp collecting view of the world, so it is another interesting, useful tool for them.
But more important, especially in the long term is that people will be able to earn their share of the multi-billion dollar search market. At the moment it is a multi-billion dollar industry and it is really only run by a handful of players. And in the same way that there are no longer just three big TV networks or three big newspapers making all the money in those media forms, we believe over time lots of experts in their area will earn money out of search.
Someone will build the best search engine around stamp collecting and earn money out of it the same way people that publish stamp collecting magazines earn money out of it, or people earn money out of a web site around it. And that is really what we want to enable and empower people to do. So it is really designed as an additional useful tool, independent of the fact that you’re making money out of it, and in the long term people really can build valuable assets out of it.
We have already seen it with specific verticals – for example, in the travel industry there are already numerous vertical search engines. You can search for things like jobs, and already people don’t use Google for jobs, or if you do a travel search you’ll end up using Expedia or Tripadvisor. People have taken these big niches and developed their own technology and it is really useful to do that. This verticalization made us say let’s enable anyone to do it around any topic, and that’s really what we are all about doing.
John: You were using an analogy of cable television or satellite TV where we have seen an explosion of niche areas both from the publishing side and from the consumer side. What is the public awareness like, and the uptake rate for this technology, and is it increasing? Have some of the recent activities of the big search players been helping you with awareness?
Grant: It is definitely helping us. Before we needed to explain “why would you bother? why create vertical search engines?” And now it is like “well the big guys are doing it, so yeah, this is a good idea.”
People inherently understand that media does get verticalized. It has happened in every other media form and it is going to happen for search now. But it is not something that people have thought about so much. So people see that Yahoo! and Google are doing it and think “let’s check out the company that has been doing it for three years.”
Eric: How do you get users on board? What makes a user seek out a custom search engine instead of just wrestling with the sometimes problematic business of working with a horizontal search engine?
Grant: I think people will continue to use horizontal search engines for a long time. The game is very, very young for vertical search. We know it is very hard to get people to do anything different. It’s going to take time, I guess, in the same way that back at the start when you had the first bloggers or the first people who had a web site, why would any one go on the internet and look at that web site? Or why would they do anything differently? But what we have really tried to do with our technology is to make it easy for people to see it and use it and see some value in it and interact with it.
For example, the way we have designed our search widget so it’s more than just a search box – it kind of shows you what is going on so that it draws the users in and then they say “wow this is kind of interesting” and then they have a bit of a play and they really look at it.
People see that when they get a swicki set up they do start to get traffic and people are using it. The main use case scenario is that people will go to one of the main search engines, type in something, find a web site and see one of these search engines and buzz clouds and use it and think “oh this is kind of cool”, and then they’ll come back and use it again and over time they will learn about it.
But will we ever get to the stage where people say “where I am going to find a search engine around mountain biking or stamp collecting or ball bearings, or whatever it is, and then use that?” That use case might not happen for a long time but I do strongly believe that it will happen in the same way that we automatically don’t use Google for doing a job search. We might use Google to find a job search engine and then use that. We are just accustomed to that because that makes sense. We’ll do that more and more.
John: You raise a kind of “meta” question here, which is how to go about finding a vertical search engine, which is itself a search task. How does Eurekster help with that?
Grant: We have a directory and a way for people to search it, and we’re seeing an increasing amount of traffic – people cruising around looking for search engines. But we really do think that most of the time it is kind of like — you know, you’ve got your blog search tools like Technorati, but lots of people who read blogs just go directly to them, they don’t have to keep going to Technorati each time for the info. We don’t want Eurekster to become a big brand so that you have to come to Eurekster to find the search engine. That might well happen, but we really want to extend the brand that people have already got. People who already trust your blog or your web site — we want you to be able to extend that into search so that people will come back and read your blog and use your search engine and whatever other bits and pieces you have to offer.
So we are about extending other people’s brands. It is not about our brand vs. the big guys, it is about tens of thousands, hundreds of thousands brands out there in the world. They can extend the relationship they have with their audience into search.
It is kind of like if you look back at the start of the web — the main publishers would have a web site with a view to try to get subscriptions to their offline magazine and now you find that people have realized that this web site thing is earning more money than the offline and it is inherently valuable as its own channel. And we think the same will happen with the search and all those brands will reinforce each other and we’re a printing press are enables people to do that really.
John: Well we certainly appreciate your perspective on the market place. Let’s get some more of your insights into the technology .Where you guys have really set yourselves apart is in innovating a whole way that a Swicki can improve over time. You have features that it real easy for the Swicki creator to promote or demote a site right from the search engine result page itself. And you also have the ability for the Swicki to learn by monitoring the popularity of various results. So these are things that are above and beyond what we are seeing from some of the other players in the space. So I guess the question is this: are users in the market place placing value on these features and are they enough to stay above the noise of the big guys getting in?
Grant: That is a good question and I guess time will tell. But the main reason we developed those features is that once you give people a flavor for the idea that they can publish their own search engine, that basically the feedback that we get is that they want more and more tools to change things. You give them a little sniff or a flavor of it and then they just want to do more and more and have more control in the same way that when you’ve got a web site, you want to be able to change this and that and do all those things. We really have added an awful lot of things that people can do, and what we have to do now is make it easier and communicate it better, and that is what we are busy working on.
But a couple of other things are really kind of key to the way that Swickis passively learn from users. We know that a very small population of the users are into tagging and actively interacting with it, but we can learn from every interaction with a search engine. What people are finding valuable and not valuable, and where they get their information, and we feed it back to the people who do care so they can improve it.
And so like I say, we have been doing this now for a long time and learned a lot about what does and doesn’t work, and we’ve got some things in the wings that will extend that further.
I am sure the bigger guys will follow these things eventually, because we started with a feature rich product and then learned from the market and developed them.
John: So to build on that, can you share with us some of the things that you have planned for upcoming releases that are going to continue to extend your position and your lead in making it easy for people to really fine tune and customize their sites?
Grant: Because we really do have some incredibly rich features here, we’ve been doing an awful lot of UI testing, and it is really about exposing those in a way that is easier for people to understand what is going on and the power of what they are doing. So it is really about improving the ways that people can interact, and funneling the appropriate information to those that can do something about it and really engaging people in something more than what you can do on a normal search results page.
Until now, we have really made the search results page look mostly like any other result page and the features are there but they’re kind of hard to find and interact with. We really want to bring those a lot more alive and extend a whole lot of community features around those, and so you will see those things rolling out within the next few weeks.
John: One topic that has caught my attention in general is reputation. Now it seems to be a major component in many kinds of social networking technology on the web. What are your thoughts on how to track and leverage the reputation of the contributors or the reputation of the Swickis themselves?
Grant: That is a key integral part of what we’ll be rolling out. What happens is when you are a Swicki owner and someone suggests you add a site, you have a little star thing that says this person’s got a high reputation – people always accept their recommendations. Or maybe they don’t, maybe they’re always spamming.
These are things we’ve worked out how to calculate based on how they interact with Swickis and these are things that we are going to be exposing. So it really is about making it easier for people to get this information and do a combination of automatic rating based on how they interact with Swickis and acceptance rates and also the more typical reputation where you could say “I like this Swicki” or “I don’t like this Swicki” or this contributor or things like that.
John: Let me do a quick follow up to something you mentioned earlier – the whole passive learning feature of the Swicki. I agree with your assessment that when people come to a search engine they are usually motivated to solve a problem and they are usually not going to spend a lot of time actively annotating or tagging results. Can you talk a little bit more about how the passive learning feature works?
Grant: At our original company Global Brain from way back in ’98, the basic idea was that you could learn from people’s interactions with search results — something about what they are doing. So if people search for something and consistently click on the 8th result and spend time there, we’ll move it to the top and we’ll move the other ones down. So that’s one level — you can change the results based on how people interact with them.
And then the Swicki idea extends that in about three or four different ways, so we can learn if you set up a filter that says this web site is important or this one is not or things like that. We can learn and then we can suggest how you might want to add this web site to the list of important web sites because people are clicking on this, but it is not included in your important list of web sites. So it’s more like add these results from here, or take this one out.
And we learn based on user interaction what terms are popular to populate the buzz cloud and that helps people to use this search more, and understand it. And we also learn where there are gaps in the search results — where people are always trying to find something but are not really clicking on anything or finding anything. So you might want to fill out the results list to make your Swicki more useful. We are really taking that learning and applying it in lots of different ways as we’re basically learning what is and what isn’t working.
We have found these learning things are actually a key way of getting input from all the people that count but that don’t take the time to tag.
John: So you are really tracking the click stream of users that come through a Swicki and I guess we are all aware of the debacle with AOL leaking searches and search terms a while back. Does the specter of monitoring what users are searching ever come up as a concern for users and how do you address that?
Grant: Yes it does and essentially it’s a key thing we have done in our architecture. Unlike the large search engines, that store even anonymous id’s with the user’s whole history — and they do do that, and that’s how AOL got the information — what we do is, for any particular time period we store the click result. And then we will aggregate the clicks on the mountain biking one with the clicks on the ball bearing one, so an individual’s search history is actually completely lost. We don’t store that long term at all, so there is no way of identifying an individual.
These things are really about what this community cares about, and so an individual can contribute to that but the unique ID is then discarded, and we want to discard it because we don’t want to have any of these privacy issues.
John: I’d like to go back to another area we talked briefly about, which is that part of the value for publishers is monetizing their Swicki. I’m wondering if the nature of the Swicki — and the whole verticalized search concept — if that changes the advertising model? For example, it would seem that you would get a lot higher click-through rate and a higher conversion rate than on a horizontal search engine. Is that a fact and how does that change the model?
Grant: That is a very good question, and the key thing I want to emphasize here is that once you get more vertical forms of media, the value of the advertising goes up, and typically it’s an order of magnitude. The costs per view in a magazine that’s targeted is typically a lot more than a cost per view in a newspaper because you know who the audience is and you basically know where they are coming from.
So again, rather than having us dictate to people what the best way to advertise to a particular targeted group is, we really want to be as open as we are with the search results. You can get results from anywhere and we want people to advertise and monetize in any way possible. So in the same way that magazines work — some of them are completely free and ad driven, some are subscription driven — there are a lot of different ways we can advertise and choose what sort of ads are appropriate.
We want to do the same thing so that is why we have a selection of different kinds of advertisement and we are basically going to be extending that and opening that up to as many ways as possible. Basically in the same way people get creative on how they can monetize their web sites, depending on their audience and what they are doing, they can do the same with their Swickis. We can see Swickis that are valuable – not dissimilar to the Yellow Pages which is actually one hundred percent ads but which is actually extremely useful for users and useful for everyone.
We can see people doing a range all the way from that through to things that have got absolutely no advertisement, and all the ranges between, like pay per click, pay per action, all the different ways. We have actually been working on some solutions to make it easier based on the fact that you have a targeted audience. We have come up with a new advertising method called preferred placement that we are learning and playing around with as well that will make it easier for people.
So to answer the question what we are working on: we are convinced that there is no one correct answer; there will be lots of different answers, and our aim is to offer people as many options as possible.
John: Is that preferred placement option available today?
Grant: It is.
John: Could you describe that a little bit — how it differs from traditional search engine advertising models?
Grant: The reason we came up with the preferred placement model for advertising is we want the advertisements to be as relevant to the swicki topic as the rest of the search results are. It is best to explain how Preferred placement works with an example of a jigsaw swicki (Editor’s Note: see http://jigsaw-puzzles-search-engine-swicki.eurekster.com/ for the example Grant is referring to).
This swicki gives great results that are all about jigsaw puzzles if you type for quite general terms like “dogs”, “cats” or “online games”. If we used normal keyword based advertising for “dogs”, “cats” or “online games”, as you would find in a normal Google or Yahoo search, the ads would have nothing to do with jigsaws. But with preferred placement, ads at the top are all about jigsaw puzzles.
The way it works is that an advertiser can choose a swicki that matches the topic of what they have to sell. This is just an example case of how www.jigzone.com could choose preferred placement on the jigsaw puzzle swicki (Editor’s note: doing a search on “dogs” on the above Swicki does, indeed, result in www.jigzone.com showing up as the top advertisement).
When jigzone chooses that option, we then do a deep index of jigzone.com, and every time they have matching content we show it at the top of the search results and highlight it. The swicki publisher can choose what price the advertisements are and who to accept.
The main advantages of this system are that, first, it’s much less work for the advertiser – no need to choose lots of keywords and bid on them all individually – plus they know they are getting an audience that is interested in their products so we suspect there will be much higher conversions. Second, it’s great for the end user – they get ads that are more relevant. And third, it’s easy for a publisher to get great, targeted coverage with just a few advertisers. In the long term, the price will go up and reward them for the niche audience that they have (just like, as we said earlier, advertising in magazines costs a lot more per view than in newspapers).
One of the nice things about swickis is that they automatically identify the most likely set of advertisers out there since they are the sites that are already getting traffic from the swicki. We will be exposing some of this information to swicki owners soon to help them target the best advertisers for their swicki. A lot of people that have created a swicki around a topic actually have some relationship with these advertisers, and this can really extend that and make it easier. We will also be making it easy for advertisers to come in and advertise on a set of swickis related to their products.
In addition to the Preferred Placement option, we will be offering a full range of advertising solutions for swicki owners so that they can monetize them in what ever way they see fit – for example, pay per click, pay per view, pay per action, subscription, etc. In the same way that there is no one correct way to earn money from a great website, there will be no one solution for every swicki, so look out for more options coming up.
Really, we can look at what the normal advertising answers are and play with them, but we are developing new tools that really suit the fact that this is a new kind of search engine and so some new tools will be valuable we think.
John: So this is all about bringing the same relevance to the ads that you bring to the search results?
Grant: Relevance yes, and without it, you are training people to not look at the ads.
John: You are also making it much easier for the advertiser – no more need to figure out 10,000 key word combinations, right?
Grant: Yes, for sure and we think it is really quite exciting from an advertiser’s point of view. Again this is hard because this is something new and different and they are so used to thinking “well what are my twenty most important key words?” and I have done well on those. And we say you don’t have to know. It works for the advertiser, and gives a better user experience and we think long term the value per click is going to be a lot higher as well.
John: Well that sounds really innovative. I am actually excited about going to try that out. I have not played around with that part of the product yet so…
Grant: It’s in its early days so you kind of got a sneak preview and we are trying to make that a lot better from our initial version, but I can help guide you through that process as well if you like.
John: Great! Thank you for the offer; we will follow up on that. Let’s sort of wind down with a few more questions that are oriented around where things have come from and where they are headed. So let me start by asking you what kind of search volume are you seeing across all Swickis at this time?
Grant: Ah, well it is getting close of a half million searches a day. And it is on a good incline over the past year, so we are pretty excited about the trajectory we are on.
John: That does sound very good. Do you have any estimate about how many Swickis there are currently deployed?
Grant: There are a little over thirty thousand have been created in a year.
John: Okay. So now let me just get a little bit of the cultural history behind Eurekster. I’m curious because these are rather interesting words. Can you tell us where the terms “Eurekster ” and “Swicki” come from?
Grant: I came up with the idea for the name Eurekster in the same position on the couch when I came up with the idea for combining social networks and search, so it was really kind of a Eureka moment. It really is a case of � the social things were all “ster” for example Friendster and Napster and those things and so it kind of made sense. And then swicki was really the case of a search engine that was like a wicki because the whole model until now has been that you can’t do anything to change the search results and we just flipped it around and said you can do anything you want. You can change it and move it around and so that is why the idea of the swicki came up. We want to trademark that term because we want it to become a thing, in the same way that people talk about a blog or a wicki or a podcast they can talk about swicki, because it is a search engine that people can interact with.
John: I guess the last question I had was about your business environment. I find it fascinating that you guys are thought leaders and innovators and yet you are set up half a world away from the epicenter of the search industry in Silicon Valley. So I’m curious – is this a handicap for staying in touch with the buzz of Silicon Valley; not only staying in touch but maintaining Eurekter’s buzz?
Grant: I actually think there is an advantage, sometimes, in not living and breathing the same thing as everyone else. Because if you are in an environment, and getting the same stimulus as everyone else, then you tend to think the same way as everyone else. And ultimately thinking the same as everyone else is not what is going to give you a bit of an advantage. So again, I would see that as an advantage. We have our office in San Francisco, so we get enough of the buzz anyway, but we tend to not jump onto the latest and greatest of this and that and just kind of look at our users and numbers and follow what we think is going to be best for where we are going and not get caught up in whatever happens to be the flavor of the month.
John: The other question is regarding hiring quality engineering staff. Do you have an advantage or do you have a challenge operating in New Zealand?
Grant: Well we actually have a huge advantage because there are really top quality universities here, and we are far and away the coolest thing in town. We also have a lot of high quality immigrants from all over the world choosing to come and live down here and they are always interested in looking for something exciting to do, so that is great for hiring too. We found back in the days of NBCi that we had thirty people here and we had one person leave the entire time, where the engineering team we were working with in San Francisco had one hundred percent turnover of staff. As far as engineering quality and hiring goes, it is fantastic.
John: Okay, in the spirit of the “Lord of the Rings”, since you brought it up, I did say that was my last question but I am going to have yet one more last question. Where do you see the search business headed over the next two to three years? The whole marriage of social networking into search that you guys have pioneered has certainly been one of the biggest innovations over the last few years, but what is the next wave; what else is coming down the pike?
Grant: Well I think it is safe to say that the verticalization of search is at a – we are at the beginning of that dream. Most people have no idea what verticalization of search is, so you are going to see that go more and more mainstream and when it does, then you will see the change in some of these ad models, and then you’ll see tens of thousands of people actually economically benefiting from the search engine industry. They can take the expertise that they have in a particular area and apply it to search and actually earn a very decent income and build a valuable asset out of it. So it is a way of helping all the creative and intelligent people in the world to participate in organizing the information on the internet and then earning their share of the revenue, so it ought to be really cool.
John: Well great stuff, Grant. Again, we are very grateful for your time and insights. We have enjoyed working with you guys a lot. We really admire your technology, and we share a lot of your views of where the market is headed. We think it is a pretty exciting place to be, so we are glad to have had a chance to work with you guys. We wish you the best.
Grant: Yes, no doubt we’ll be talking again at some stage and it has been a pleasure. I mean you guys — clearly from your train of thought and questioning — are very much on the ball, and it has been fun.
We have published an analysis of this interview in our blog.
About the Authors
John Biundo is a Principal Consultant, specializing in SEO/SEM, at Stone Temple Consulting.
Eric Enge is the Founder and President of Stone Temple Consulting (STC). STC offers Internet marketing optimization services, including SEO, Social Media and PPC optimization, and its web site can be found at: https://www.stonetemple.com.