Q&A With Garrett Camp, Founder & Chief Architect, StumbleUpon
Search engines are great tools to help you find things you suspect exist on the web, but they leave a lot to be desired when it comes to discovering the unexpected or unknown. Sure, you may find something really cool after doing the result-page/back-button tango for a while, but that’s an inefficient approach.
Enter StumbleUpon, a service that lets you “channel surf” the internet, displaying sites that others have “stumbled upon” and recommended. StumbleUpon is a hybrid: Part directory of human selected sites, part search engine (the millions of pages recommended by “stumblers” have been fully indexed and categorized behind the scenes), part social bookmarking service. It’s all of these things, and yet offers a unique web discovery experience unlike any other I’ve seen.
In his first “Let’s Get Social” column, Forget ABCs – The Social Media Alphabet Is DNRS, Neil Patel introduced StumbleUpon. In today’s Q&A, I asked Garrett Camp, StumbleUpon’s Founder and Chief Architect, to take us behind the scenes of this intriguing discovery engine.
Q: You started StumbleUpon in November 2001 – a pretty dark time when lots of web sites & online businesses were crashing and burning. What inspired you to start at that time? And why did you see a need for this type of service?
We realized that search engines are useful when you know exactly what you are looking for, but for discovering new and personally interesting online media keyword search wasn’t always the best option. We created StumbleUpon so people can “stumble upon” sites that have been submitted and rated by like-minded people, rather than presented with the most popular sites for a given keyword. We designed StumbleUpon to blend collaborative human input with machine learning techniques, so users could discover great content they wouldn’t have thought to search for. We’ve been focused on social media and collaborative discovery long before the phrase “Web 2.0” became popular in 2004, and now have over 2M members discovering and sharing interesting web content.
You’ve said that “stumbling” is more a process of discovery than search—you’ve even called StumbleUpon a “remote control” for the web. Can you describe how StumbleUpon resembles “channel surfing” and how you customize the experience for each user?
Search is great for textual content or if you know exactly what you want, but if you’re simply looking to browse through relevant and high-quality web content we think the “channel-surfing” metaphor makes more sense. Instead of having to search, scan, click, go back, and repeat, you simply surf directly between relevant content which has been given the “thumbs-up” by friends and other web surfers with similar interests. We have nearly 500 topics that users may choose to indicate preferences, so your recommendations become precisely tailored to your interests.
Essentially we combine community reviews and social feedback with clustering and evolutionary methods to create a personalized recommendation engine. This helps people discover interesting websites, photos and videos recommended by their friends and like-minded people.
You’ve got about 9 million sites indexed—how dynamic is that index as new content is added to the web or old content disappears?
We like to think we have the best of the web in our database. Our 2 million registered users stumble around 5 million times a day, so we have a pretty active user base. If they find something new, it’s incredibly easy for them to submit it to us. All they need to do is click the thumbs-up button on the toolbar and it’s submitted to our database. We get over 16,000 new URL submissions a day – all new and unique content endorsed by our members.
I’ve been skeptical of many aspects of social search. StumbleUpon incorporates a number of social search techniques (thumbs up-or-down voting for web pages, tagging, collaborative filtering and so on), and yet also does a lot of algorithmic heavy-lifting in the background. What kind of indexing and classification are you doing in addition to observing user behavior?
We have a classification engine which automatically places content into one of 500 predefined categories based upon on-the-page factors. This means most content submitted can be distributed to interested members even before tags have been applied. We also have the entire database indexed so you can stumble by keyword if you want to get fairly specific… for example stumbling through just pictures of “sunsets” or discovering new “fonts”. We use classification and indexing in combination with social networks and user ratings to improve the quality of recommendations and coordinate the distribution of relevant content.
StumbleUpon offers a lot of community features. You can browse the StumbleUpon community to find users with similar interests and even become “friends” with them. You also encourage people to share a lot of information about themselves. Does the user experience change depending upon how much self-info you’ve provided, or the network of connections you’ve made? How?
Your stumbling experience improves as you get more involved in the community. Once you add a picture and start reviewing pages, you’ll start meeting other community members who like similar types of sites. We’ve found that most active members will upload a picture or fill out their favorite music, movies etc… so they can start meeting like-minded people. Over time the social aspect of StumbleUpon makes it more compelling, since you’ll discover pages your friends have recommended and learn about their tastes as you stumble.
What about people who aren’t comfortable sharing personal information? What kind of privacy measures do you have in place?
We have a few privacy features, and how much information you share is completely up to you. If you want to remain an anonymous member without a public presence simply don’t upload a photo or fill in your profile. You can also choose to not share your ratings if you like, but we’ve found the vast majority of people don’t mind other people visiting their profile or reading the reviews they write.
One of my favorite features of StumbleUpon is the automatic addition of StumbleUpon ratings and a link to reviews in search results from Google, Yahoo, Ask, AOL and Microsoft. This adds a “second layer” of relevance to the results, showing you at a glance which pages have been given a thumbs-up or good reviews by stumblers. The StumbleUpon toolbar also lets you search by keyword, but sends you directly to a web page rather than showing search results. Why take that approach rather than showing a list of the top pages StumbleUpon has indexed for a particular keyword?
When developing the Search Reviews feature, we figured that most people didn’t want to switch to a different search engine, and would rather just have social aspects added to their current search engine of choice. So by adding links to Google, Yahoo et al, you can still query a multi-billion-page index, and get additional social metadata when simple queries are submitted. This means you don’t need to change your search behavior at all to enjoy the benefits of community endorsements when they exist. As for Stumbling by keyword, it’s really to just keep the navigation experience consistent, and maintain the sense of serendipity.
We’ve seen small but growing traffic at Search Engine Land from StumbleUpon users. Any recommendations for search marketers wanting to tap into the StumbleUpon community?
StumbleUpon has a unique business model that works well for marketers where we can deliver traffic directly to your site. You can target by category, age, gender and location. So for product launches, distributing audio/visual content or just getting feedback on your blog, StumbleUpon often works better than PPC approaches since targeting is precise and no click through is required. More information on opportunities is available on our advertisers page.
Talk about your future plans.
We are always looking to improve the user experience and the technology behind StumbleUpon. We are currently focused broadly on both social search and video discovery, as well as algorithmic improvements to our recommendation engine. We are also re-designing our entire site and investigating bringing StumbleUpon to other platforms beyond the browser. Stay tuned…
Thanks much, Garrett.