Greg Linden is founder of Findory, and author of the popular blog "Geeking with Greg". Findory is a personalization and recommendation service that helps users find news stories and blog posts of interest to you. Findory also has a search engine built in that personalizes its results based on your past searches and clicks. Think Memeorandum combined with Google personalized search, powered by a collaborative filtering engine. It is amazing.
I learned of Findory about 4 months ago when I started seeing traffic from Findory to my blog. I didn't have the time to experiment with it then, but recently devoted some time to understanding what it is and how it works. You should try it...definitely worth the time. What better way to learn than an interview with the founder?
Here are my 10 11 questions with Greg Linden.
1. What were you doing before Findory? Is this your first start-up?
I was at Amazon.com from early 1997 to 2002. I wrote the recommendation engine used by Amazon.com and later led the technology team that developed Amazon's personalization systems. Before Amazon, I was in graduate school in the AI group in the University of Washington Computer Science department. After Amazon, I was at Stanford Business School in the Sloan Program. This is my first start-up.
2. When did you come up with the idea for Findory? What inspired you?
I wanted to personalize information. Amazon.com has demonstrated the value of personalization in e-commerce. I want to bring personalization to information. I want to help people deal with the overwhelming flood of information in their daily lives by helping people focus on what is relevant and useful.
3. How long did it take you to build the first prototype, and ultimately bring it to market? How long have you been live?
Findory launched early with what might generously be called an alpha in January 2004. At that point, it was news only. Weblogs were added in June 2004. The site has grown and continued to be refined over the last two years.
4. What exactly is Findory? It seems to be part search engine, part recommendation engine, and part meme tracker that finds the most popular news stories and blogs. How would you describe it?
Findory is primarily a recommendation engine.
When people come to a website like Amazon.com or Google, sometimes they know what they want right away. Then they use search. When they don't know what they want, when there is too much information out there, personalization can help. Personalization learns what you want from what you do. Using personalization, Findory helps focus your attention and surface things that you may not have been able to find on your own.
Personalization and search complement each other. Findory offers both.
5. How do you compare Findory to Memeorandum, Wink, Digg, Jookster, and others?
I don't see much in common with those sites. Memeorandum combines inlined tracebacks with link popularity. Digg combines a voting mechanism for finding most popular articles with a discussion forum. I am not familiar with Wink or Jookster, but they appear to be web search engines, Wink focusing on tagging, Jookster focusing on social networks.
Findory is quite a bit different. Unlike Digg or Jookster, there is no explicit voting or listing friends. Findory's personalization learns from what you do, finds other people with similar interests to yours, and shares what they have found, all implicitly, all anonymously, all without any effort. Unlike Memeorandum, Wink, and Digg, every reader sees a different page on Findory, each page personalized to each person's interests.
6. Does Findory present a different personalized page to each user?
Yep! That's the idea!
7. I noticed Findory keeps track of all my searches and stories read for both News and Blogs. Do the recommended stories and search results get more focused as my user history grows?
Yes. Findory learns very quickly and can make recommendations even if you have only read a couple articles, but the recommendations will be more detailed and comprehensive if you have a longer history on Findory.
8. How does Findory factor in choices other users have made to the results I see? It feels like collaborative filtering. Is that another element of the secret sauce?
Findory recommends articles between Findory readers. When you read articles on Findory, Findory looks for other readers who might have the same interests as you, and then shares the articles they found.
It is a little like social networking sites where you explicitly list all your friends, then explicitly share things you found among your friends. But, with Findory, friends are found for you automatically and anonymously. With Findory, the sharing is done quietly and implicitly, all with no effort. Just read articles, that's it. Findory does all the work for you.
The personalization techniques used by Findory fall loosely into the class of collaborative filtering algorithms. However, naive collaborative filtering has well known quality, performance, and scaling problems. Findory creates fully personalized pages in real-time, works even if someone has read only a few articles, learns immediately from new data, and can scale to millions of users.
9. According to Alexa your site traffic is right up there with the other popular meme trackers? Who do you consider your closest competitor it terms of features and functions?
Findory has been growing at between 15-30%/month since launch. Findory's current traffic is about the same as Memeorandum, Rojo, and MSN Start.com
To my knowledge, there are no other start-ups doing personalized information like Findory. However, the search giants have made tentative efforts toward personalization. Google has an experiment with "recommended stories" on a small section of the Google News page and also has an active effort in personalized web search. MSN has experimented with personalization of a small section of the MSNBC and MSN Newsbot pages. Their efforts use different techniques. They do not adapt as quickly as Findory nor offer as fine-grained, targeted, comprehensive personalization.
10. What would you like Findory to be in the next year or two? What are users asking for?
Findory continues to grow rapidly. Over the next two years, I would like to add to our early personalized web search, offer additional advanced customization features, expand our news and blog crawl, build our own advertising network, offer versions of Findory in non-English languages, and launch in other areas such as videos, podcasts, and images.
11. What is your business model going forward?
Findory has its own personalized advertising engine layered on top of Google AdSense. Unlike AdSense or other contextual advertising engines, Findory's advertising is personalized. Findory individually targets ads based on both the content of the page and the articles each reader has read. As of December 2005, Findory is cash flow positive.
I have been using Findory for a week or so. Initial results are impressive. I plan to use it for several weeks to see how the results get more focused and personalized over time. Maybe I will do a follow up post on my findings.
Have you used Findory? What is your experience?
I use Findory daily, and have found that it really does a good job learning about topics that interest me. It learned quickly, and after several months of using it, I can still see that its algorithms are learning about me and further fine tuning the personalization. I'd recommend Findory to anyone, if that tells you anything. :)
Posted by: Otis Gospodnetic | February 09, 2006 at 12:04 AM