SUBSCRIBE NOW!

Vol. 12 No. 9 — October 2004

However, those of us who spend a great deal of time looking at new technologies know that many exciting ideas and useful products come from small companies or sole proprietors. Often these small companies offer services long before the "big guys" catch up, if they ever do.

Matt Wells of Gigablast [http://www.gigablast.com] is an example. His general Web search engine continues to take major steps toward becoming a major player on the Web search scene. The same goes for Rick Skrenta and his small team, who build Topix.net [http://www.topix.net], one of the most useful browse/news search tools that I've seen in a long time. It's easy to use, offers syndicated feeds, and contains news sources not found in other databases.

Personalization on the Web (results based on user needs and interests) gets plenty of attention these days. We've seen personalization of results projects from Google Labs, a personalized news tool from Microsoft called Newsbot, and many others, including Amazon.com's personalization of book pages with other information you might find useful.

One company, Findory, already offers an online personalized news service. This Seattle-based company is led by Greg Linden. Before starting this company, Linden worked at Amazon.com in the personalization division.

Recently I had a chance to conduct an e-mail interview with Findory's Greg Linden.

Greg, what is Findory News?

Findory News [http://findory.com] is a personalized news site. It learns your interests from the articles you read, searches thousands of news sources, and helps you find the essential news of the day.

Why would someone use Findory News?

Findory helps you find what you want faster and more reliably. All you have to do is read news. Findory learns what you like and the front page rearranges itself to match your interests. By searching thousands of sources for news that interests you, Findory News helps you find articles you might otherwise miss, while still providing broad coverage of recent major news.

Personalization seems like a hot topic these days. How do you define it?

I define Web personalization as delivering different and unique content to each individual customer based on the customer's interests. It's your own version of the Web site, a site just for you.

To distinguish personalization and customization, I would argue that personalization uses implicit interests and customization uses explicit interests. Personalization learns what you like from your actions; you are what you click on and what you buy. Customization requires you to explicitly specify what you want; you are what you say you are. My Yahoo! is an example of customization. You tell the site what you want. Amazon.com is an example of personalization. The site learns your interests and adapts.

In a world with a glut of information, personalization offers a way to find focus. It doesn't waste your time showing you what everyone else sees. It learns what you like, shows you what you want to see, and filters out the rest. That's personalization.

Tell us about your background and why you developed Findory. Where did the idea come from? How long did it take from idea to launch?

I started at Amazon in early 1997, helping build what was a small online bookstore into a shopping site for books, music, video, kitchen, computer games, electronics, hardware, and just about everything else you can imagine. Amazon is well-known for having the best online personalization systems on the Web; the site learns your interests and helps you find what you want from Amazon's massive catalog. It's a store for every customer, as Jeff Bezos likes to say. I was fortunate enough to be building Amazon's personalization system from the beginning, from writing the first recommendation engine used by Amazon.com to later leading the software team that built all Amazon's personalization systems. Before Amazon, I was in the graduate computer science program at the University of Washington studying artificial intelligence. Just before starting Findory.com, I was at Stanford University in the Sloan Program at the Graduate School of Business.

Findory.com started in mid-2003 developing personalized search technology. The early prototypes we developed analyzed user search and click behavior and modified Google search results based on the user's history. The technology was particularly promising when someone was repeatedly refining their search — adding, deleting, or changing terms — trying in vain to find what they want. Unfortunately, we required large amounts of data on user search behavior. As a startup, Findory.com didn't have this data, nor were we able to acquire the data or find a partner who would allow us access to the data.

At this point, late 2003, we began experimenting with personalized news. Personalized news is not a new idea. People have been working on it for many years. But it is quite challenging to do well. News is perishable — old news is no news — and the rapid expiration of the data thwarts many algorithms that could be applied to the task. Previous attempts have mostly either been keyword-based, essentially finding articles with the same keywords in them, or category-based, simply finding articles in the same broad subject category. Both methods produce mediocre results, showing articles that are too specific or too general, not at all compelling. However, we found a variation of the techniques we had developed for personalized search produced high quality results for personalized news. Following the start-up maxim of "launch early and often," Findory News was launched as a beta in January 2004. After several iterations on the algorithms and design, Findory News came out of beta in March 2004.

Aside from building the technology, what else goes into launching a search company? Is it hard getting attention in a time when the big guys seem to get a bulk of the ink?

Start-ups have some advantages over established firms. They're nimble, creative, tolerant of risk, and have no legacy systems to maintain or older products to cannibalize. But they have many disadvantages as well. They're short of resources, have an unknown brand and product, no existing user base, and have difficulty being noticed.

I've found PR and marketing to be [among] the bigger challenges for Findory.com. Given that Findory is seed-funded, we don't have a marketing budget that can support large advertising campaigns. While Findory is doing some focused, targeted advertising, we rely on word of mouth and positive press to increase awareness of Findory News. Much of the PR work is the normal routine of contacting and talking with reporters, but Findory has also sought to take advantage of the increasing popularity of alternative press, especially popular Weblogs. Weblogs do have an interesting network effect, where an article in one Weblog may spur a conversation that spreads across the Web, discussions held over many days on different Weblogs.

Can you tell us a bit about the technology that Findory uses? Was it developed in-house? What does Findory look at to help it personalize results? Do you do all of the coding yourself?

Findory's technology was developed in-house. It primarily uses statistical analysis of user behavior supplemented with analysis of article content. Findory has filed for two patents on the technology.

How is Findory funded?

Findory is seed-funded by the founder. At our projected burn rate, the funding is sufficient into 2005.

Can you tell us about your relationship with Topix?

I've been in contact with Rich Skrenta for several months. Topix.net and Findory.com have a cooperative relationship. While we are trying to solve similar problems — both trying to provide focus into the overwhelming amount of news on the Web — we have very different solutions. Topix.net is using fine-grained subject categorization that provides readers a way of reading articles on very specific topics. Findory News takes a different approach: Just read articles. We'll learn about your interests and help you find other articles that interest you. The approaches are complementary, so we have no reason not to cooperate with each other.

Topix.net has been kind enough to allow us to use some of their news articles. Although Topix.net data represents only a small fraction of the news in our database, the breadth of the Topix.net news crawl does improve the quality of Findory News. In exchange for the use of their data, we credit Topix.net on Findory News and drive traffic to their site.

Does Findory also crawl Web news sources? How many sources does Findory contain in its database? Do you have any plans to increase this number?

Findory does do its own Web crawl. We currently have over 2,000 sources in our news database. We're always seeking to expand our news coverage.

Where do you see the company in 5 years? What services will Findory be offering that are not available today? Any plans to add any fee-based services?

Google's first product was in Web search, and later they launched Google News. Findory.com's first product is Findory News, but later we will launch personalized Web search. As Findory News grows, we will have the traffic to be able to launch our personalized search engine. Using the same strategy as A9 [Amazon's Google-fed search engine] and Vivísimo of building on top of existing search engines, Findory's personalized search doesn't require building a massive search infrastructure, just sufficient traffic to drive the algorithms and analysis.

Findory's personalization technology provides focus anywhere there is a glut of information. In 5 years, Findory will be the personalized portal, providing personalized searches over Web pages, news, products, and anything else you want to find on the Web.

Findory does plan to add targeted advertising, but [has] no plans for any fee-based services. Our targeted advertising will be text-only similar to Google AdWords, but use the full history of user interaction with Findory rather than just the current search keywords. The advertising will be unobtrusive, relevant, and useful.

Have you had a chance to take a look at what MS is doing with their Newsbot tool? How is Findory different?

The beta version of the U.K. version of MSN Newsbot has gone through several iterations. At first, MSN Newsbot required users to sign in using MSN Passport and had a lengthy delay before the page became personalized. More recently, MSN Newsbot has moved closer to Findory News, offering immediate personalization without sign in. Some

differences remain. The quality of the personalization on Findory News seems to be substantially higher; as I said before, it's tricky to get personalized news right. Second, the entire Findory News site is personalized. MSN Newsbot seems to limit its personalized content to a few articles in the upper right corner of the screen.

I'm pleased that MSN has chosen to develop a personalized news site. It validates the idea of personalized news and attracts attention to Findory News.

Google Labs launched a personalization tool recently. What do you think of it? How about what Udi Manber and his team are doing at A9?

I'd say that Google's search personalization is more customization than

personalization. It's a profile-based technique, requiring users to fill out a lengthy form about their interests before they can get started. Google seems to prefer this type of approach because, as Craig Silverstein (Director of Technology at Google) recently said in a ZDNet interview, "the computer only has to be intelligent [enough] to use that information to try to help you out" rather than also trying to be intelligent enough to learn your interests. I have a huge amount of respect for Craig, but I think he's missing some key points here. When you rely on people to tell you what their interests are, they 1) almost always won't bother; 2) if they do bother, they often provide partial information or even lie; and 3) even if they bother, tell the truth, and provide complete information, they usually fail to update their information over time. But, requiring an explicit profile isn't the only reason why Google's personalized search doesn't work well. It's also targeting the wrong problem. When people search, they're on a mission. They're looking for something. If you want to help me in my search for, say, a "leather sofa," it isn't enough to apply a broad profile that says that I tend to be interested in furniture. You need to pay attention to what I'm doing right now and use that information.

A9 doesn't do much yet other than track your search history. But that is the first step in personalized search. And I think it's pretty clear that A9 intends to pursue personalized search. Udi Manber has made several statements about it, such as in his recent InternetNews interview, where he recognized that "relevancy is different from user to user."

Personalized search is inevitable. With only one general relevance rank, it is increasingly difficult to improve search quality because not everyone agrees on how relevant a particular page is to a particular search. At some point, to get further improvements, relevance rank will have to be customized to each person's definition of relevance. When that happens, you have personalized search.

You recently launched an e-mail and RSS service. How is that going? Do people still find e-mail alerting tools useful? Do you think syndication is going to be as big as many believe it will be?

Your Daily Findory News, the Findory News e-mail service, and Findory News RSS feeds are currently in beta. Both are doing quite well. The daily e-mail is like getting a front page made just for you delivered on your doorstop every morning. Our readers seem to like the feature, and it encourages people to come back to Findory News every day.

Findory News RSS feeds provide an alternate channel to get to Findory News. To my knowledge, Findory News is the only site offering personalized RSS feeds, feeds that differ from user to user based on their interests. Not only does it allow readers to see their personalized Findory News in their favorite news reader, but also it allows bloggers to feature their personalized news — the news they see on Findory News — on their Weblog for all their readers to see and enjoy. It's a great service.

Syndication is interesting. It makes news even easier to access than it already is. But news already is pretty easy to access on the Web. The real problem is focus. The question is no longer, "How do I get information?" The question is, "What information do [I] need to know?" Power users of RSS probably have encountered this problem already. When you have 5­10 RSS feeds in your RSS reader, it's great. When you have 100, it's not. You have long, unaggregated, unprioritized lists of news to wade through every day. Unlike RSS readers, Findory News provides focus. Rather than organize your news by source and listing by date, Findory News aggregates and prioritizes all of your news by your interests.

Is Web search a zero-sum game? Can niche tools survive as the big guys offer more and more services?

The Web search market is growing rapidly and people's thirst for information and knowledge seems unbounded, so I don't see how Web search could be seen as zero-sum. This isn't even close to a mature market.

Niche tools will thrive, since specialized tools and databases can often beat the performance of generalized tools and databases for at least some class of searches.

If tomorrow you were given the capital and a pool of great developers to build the perfect search tool(s) ... Would you decide to go after Google and Yahoo! or build niche search tools? As an observer of the industry, what could the "big guys" do better?

I think the biggest opportunity lies in personalized search. At some point, a generic relevance rank just can't get any better. It's as relevant you as can make it to the average user. When you recognize that different people have different perceptions of what is most relevant to them, then you have a huge opportunity to improve search. This is both where the big guys — MSN, Yahoo, and Google — could do better and [is an] opportunity for Findory.com.

Findory doesn't require registration. From a business angle why did you make the decision not to learn more about your users? Have users expressed concerns about privacy? Would knowing more about a user aside from what they read help improve the personalization?

Findory News doesn't require registration because it doesn't have to. I want readers to have to do nothing other than read news. That's it. The site helps them find other news of interest, but doesn't get in their way. It's easy, simple, and useful.

No one has expressed concerns about privacy, but that's probably because Findory News knows nothing about you personally. It doesn't know your name, your address, your credit card, or anything else. Readers are linked to a random, anonymous number, but the system doesn't know that any particular number is actually Greg Linden or Gary Price.

Knowing more might help a little, but not as much as you might think. Findory's personalization technology is optimized to work well even from small amounts of data. It can make a rough guess at your interests after you've read just a few articles. More data allows it to refine its guess, smoothing out the edges, but the guess was probably fairly close already, so the improvement is incremental.

Does Findory have any plans to license its personalization technology?

We've talked to a few firms about licensing the technology, but we're focused on developing our core product, Findory News, and have been resistant to deals that might distract us from that mission.

Thanks for your time and good luck with Findory.

Editor's Note: Since we did the interview Linden launched another site, Blogory [http://blogory.com]. This offers the same type of personalized content that you'll find at Findory. The difference is that Blogory provides personalized content from Weblogs.

Also, MSN Search launched its MSN Newsbot in the U.S. [http://www.info
today.com/newsbreaks/nb040802-1.shtml]. Greg shares a few comments about this Microsoft project on his Weblog [http://glinden.blogspot.com/2004/07/msn-newsbot-review.html].