StumbleUpon Revs Forward After Exiting eBay; Rivals Facebook As Social Discovery Engine
For a company that’s all about helping people wander the Web and make serendipitous discoveries, San Francisco-based StumbleUpon has been on a remarkably straight path since 2002.
That’s not to say the venture-backed company hasn’t gone through big changes. It has: from tiny Firefox add-on maker in Calgary, Alberta, to angel-funded SoMa startup, to eBay subsidiary, and finally (in 2009) back to being independent.
But throughout those changes, StumbleUpon’s technology and business model have stayed remarkably simple and consistent. Turn on the company’s browser extension or Web bar, and you’ll see a “Stumble!” button that will take you to a semi-random website or video chosen to match your interests. You have the opportunity to give each site a thumbs-up or a thumbs-down, share it with others, or write a review—information that helps StumbleUpon’s collaborative filtering algorithms improve future recommendations for you and people with interests similar to yours. Every 20th stumble will take you to a sponsored site; publishers pay StumbleUpon 5 cents per visitor.
That’s it—that’s StumbleUpon’s whole business, and always has been. Of course, the company has introduced variations such as StumbleThru—a way to discover content within specific sites like PBS.org or Wikipedia—and it’s got nifty iPhone, iPad, and Android apps for stumbling on the go. But the only dramatic change at StumbleUpon, apart from its short-lived fling with eBay, has been the growth of its audience.
Today the service has 12.3 million active users who stumble about 600 million times per month. As a source of social-media traffic to sites around the Web, StumbleUpon is second only to Facebook—it accounted for 33.4 percent of all visits from social-media sites in October, compared to Facebook’s 43.8 percent, according to statistics from Web analytics firm StatCounter. StumbleUpon far outranks Twitter and social bookmarking sites like Digg, Reddit, Delicious, Slashdot, or Y Combinator Hacker News. Yet for all that power, StumbleUpon is rarely mentioned in the same stories with media darlings like Facebook—which has led ReadWriteWeb’s Richard MacManus to call the startup “the silent social media success story.”
Part of the reason for the company’s relatively low media profile may be that few journalists (at least among the ones I know) have time for undirected skimming. StumbleUpon is definitely not the tool to use if you’re in goal-directed research mode—you stumble when you’re in the mood for something new or unexpected. “It’s not a replacement for search engines,” says StumbleUpon co-founder and CEO Garrett Camp. “It’s for when you want to explore and see what’s out there.”
True, you could always turn to Google or other search engines for this purpose. If you’re daydreaming about where to go on your next vacation, for example, you could try a vague search query like “cool places to go.” But if you do that, Camp argues, there’s roughly a one-third chance that the sites you find will be blogspam, whereas every site in StumbleUpon’s database is either filtered by other users or sponsored. (Even sponsors’ sites get filtered out if enough users give them a thumbs-down.)
You can also lean on Facebook friends as a kind of proxy discovery engine, and many people do, as StatCounter’s statistics show. But link sharing isn’t what social networks were really designed for, Camp argues. “On Facebook, people share stuff and are starting to ‘Like’ stuff, but it’s not the most common activity,” he says. “The fact that Facebook has 500 million users but we send websites almost as much traffic as they do indicates that people are sharing and discovering about 30 to 50 times as much on StumbleUpon as they do on Facebook.”
There is, however, an important social element to StumbleUpon: you can follow other users and see what sites they like, and they can follow you. In fact, if you scan your Twitter, Facebook, or Gmail contact lists and connect with those same people on StumbleUpon, it will greatly improve what engineers call the signal-to-noise ratio—in this case, the proportion of StumbleUpon recommendations that match your own interests. That way, says Camp, “You get the best of the social graph and the collaborative filtering graph. When you put those vectors together, you can’t get that kind of information density anywhere else.”
With users so committed that they’re stumbling hundreds of times per month, it’s no wonder advertisers are attracted to StumbleUpon. “If you’ve got really good stuff, you can use StumbleUpon to seed traffic,” says Camp. “Every viral advertiser you can think of its out there using StumbleUpon. We’re not allowed to name names, but pick a big site that calls itself viral, and there is an 80 percent chance that they are using StumbleUpon ads to do it.”
As a privately held company, StumbleUpon doesn’t release financial data, but Camp says that the company is breaking even at the 5-cents-per-visit level. That’s less than the pay-per-click costs that many advertisers are accustomed to shelling out for Google AdWords ads, so StumbleUpon may have room to increase its rates. “Once we up the price we are going to have a huge margin,” Camp says.
How StumbleUpon’s model gained so much momentum—and pulled Camp and his co-founder Geoff Smith out of their Calgary student digs and deposited them at the center of the Bay Area’s Web startup culture—is mainly a story of luck, skill, and persistence. When I met with Camp last month, he broke the tale down into four phases.
Phase 1 took place in Canada in the fall of 2001. Camp had just started studying for a master’s in software engineering at the University of Calgary, and was interested in collaborative systems and recommendations. Thanks to his government stipend, he had some free time, and he began working with Smith—the roommate of a childhood friend—on a collaborative filtering extension for the then-new Mozilla Firefox browser.
A prototype was ready by February 2002. “We were the prime example of getting into the Mozilla platform early,” Camp says. “I think we were the 138th extension for Firefox ever written, and we got a top rating very early. People liked it, and we followed the comments and fixed anything that was wrong. After a while we had 4.8 stars out of five, and we started to get thousands of installs a day, for free.”
This phase continued for nearly three years. With co-founders Eric Boyd and Justin LaFrance, Camp and Smith incorporated StumbleUpon, but the team didn’t have the resources to market the service or build up the advertising model. Camp says his share of StumbleUpon’s revenues, together with his stipend and his teaching-assistant wages, barely covered his rent. “When you have no cash or employees or office, it feels more like a project than a company,” he says. “We were just Alberta-numbered-company-such-and-such doing business as StumbleUpon, and that was it.” But the crucial number—the size of StumbleUpon’s user base—did continue to grow.
Phase 2 began when Camp visited Stanford for a conference in September 2005. There, he met Brad O’Neill, an individual investor and executive at data clustering software maker PolyServe who’d made money in PolyServe’s sale to Hewlett-Packard. O’Neill liked StumbleUpon—which, at that point, had about 600,000 users—and he began introducing Camp to other Silicon Valley investors, including Ron Conway, Ariel Poler, and Ram Shriram. “I’d never raised money before. I’d never even done a PowerPoint before. I was totally winging it,” says Camp. “But when they saw that we had no money, no office, they were impressed with how far we had gotten.”
In the end, StumbleUpon raised $1.2 million in seed funding on a valuation of “4-something million,” according to Camp. That enabled the company to move to an office in San Francisco’s South of Market district in February 2006. In light of the pre-money valuations that some early-stage startups like Quora are getting away with these days, Camp says, he should have held out for something higher. “If I had that traction today I would ask for a pre-money valuation five times that,” he says. “But at the time, we were a couple of kids from Canada, and it seemed like a great deal.”
But more money was on the way. 2006 was a whirlwind year: almost as soon as Camp and Smith arrived in San Francisco (Boyd and LaFrance had left the company by that time), they started taking meetings with big Internet players like Google, Yahoo, AOL, and eBay, all of whom expressed interest in a potential acquisition. By late 2006, “we started to get informal offers—not term sheets, but strong verbal interest,” says Camp. “But until eBay, it never seemed like the right fit or the right price or the right time.”
To keep growing, Camp and Smith knew they’d have to either raise venture funding or be acquired. They were engineers at heart, and they didn’t relish the idea of spending lots of time as supplicants on Sand Hill Road. “When eBay made a really good offer—which meant we could stay in our office and do our own product engineering stuff and not have to do all the fundraising—it seemed too good to be true,” Camp says. StumbleUpon accepted the offer, and the eBay acquisition closed on May 30, 2007. (The exact amount of the offer was never disclosed, but O’Neill’s LinkedIn profile states that it was “over $75 million.”)
And so began Phase 3, the eBay years. With the auction giant’s resources, StumbleUpon was able to hire enough engineers to complete a much-needed site redesign and, just as important, to introduce a Web-only version of the Stumble button, so that users wouldn’t have to download a special browser extension to use the service.
But it quickly became clear that eBay’s offer really had been too good to be true. “It’s nothing that’s their fault, it’s just that it’s a big company, and I didn’t have any big-company experience,” says Camp. “During the negotiations, there were so many different points I hadn’t necessarily thought about, such as the finer points of post-integration dynamics.” For example, Camp says it became more and more difficult to get the management approval the company needed for new hires and engineering changes, especially after the 2008 recession led to company-wide hiring slowdowns and even layoffs. “There was a list of probably five to 10 things that we really needed to do, and I remember realizing that it would be very difficult to make all those changes and get back to the agile situation we were in before,” Camp says. “So at some point we suggested a spinout.”
It was a good moment for such a suggestion. John Donahoe, who succeeded Meg Whitman as eBay’s CEO in March 2008, seemed more focused on eBay’s core auction marketplace than Whitman or founder Pierre Omidyar had been, Camp observes. And discussions were already underway over the potential spinoff of Skype. (In late 2009 eBay sold 70 percent of its stake in Skype to a consortium of venture and private-equity investors.)
“We just got lucky with the timing,” says Camp. “We suggested it at the same time they were doing the other spinoff, and we were able to convince eBay executives this would be a good idea. We also got help from Ram [Shriram]—the first term sheet came from him. It was a pretty complex negotiation, but finally we got to a permutation that everybody was happy with.” In April 2009, eBay sold StumbleUpon to a group of investors that includes Shriram’s Sherpalo Ventures, Accel Partners, August Capital, and Camp and Smith themselves.
Which led StumbleUpon to Phase 4—independent again. Today the company’s 58 employees occupy offices just a block away from the startup’s original SoMa location. They’ve been busy building things like the mobile apps and a custom URL shortener (called su.pr), shoring up the service’s backend infrastructure, and improving the self-service advertising portal, which lets advertisers sign up for pay-per-visit traffic directly. (Up until 2006, the company obtained advertisers through other networks such as Commission Junction or ePilot.)
The Web-only browsing option developed at eBay turned out to be a very good bet: today one-third of stumbles happen that way. Another third come via the Firefox extension, and the last third are through add-ons for other browsers such as Chrome and Internet Explorer, plus the mobile apps.
Camp argues that StumbleUpon’s paid-inclusion model works better for advertisers than other forms of traffic generation, such as Google or Facebook ads, because the company is better at choosing sites users will like—and, just as important, at steering them away from sites they won’t like. “We collect thumbs-up and thumbs-down information for every site, and people rate ads as much as they do other content,” he says. “So we can easily figure out if an ad is good or bad, and if it’s bad we can deprioritize it.” Within the StumbleUpon ecosystem, moreover, there’s always the possibility of free exposure—what the brand advertising world calls “earned media”—if users actively share the paid sites with their followers or others.
Very few StumbleUpon users complain about the sponsored content, according to Camp. “If our algorithm works well, you will find it interesting,” he says. Some of the company’s top advertisers “are just publishers who want to drive traffic to their microsites or blog posts,” Camp says, citing Intuit’s Mint.com as an example.
Activity on StumbleUpon has doubled every year since the company started, and now exceeds 6 billion stumbles per year. “It’s getting crazy,” Camp says. “We are going to be at 100 billion in a couple of years.” Which means the company’s engineers are working hard implementing distributed databases and the other key technologies that will allow the company to scale up its backend to keep up with the growth.
A year from now, the company could have 20 million users, 40 million, or even more, Camp predicts. “If you look at Facebook, they hit the hockey-stick part of their curve in 2007 when a couple of things came together,” he says. “We don’t know when that is going to happen for us. Maybe 2011. So we’re trying to get ready.” In the end, it might just be serendipity that rules StumbleUpon’s fate.