Optimizely’s Assault on the One-Size-Fits-All Web

5/31/12Follow @wroush

Dan Siroker has a habit of looking at the future and imagining something very different from the present. Back in 2007, when George W. Bush was still president, he left his plum product-manager job at Google to work for a junior Illinois senator named Barack Obama, as director of analytics for the candidate’s presidential campaign. We all know how that turned out. Today, Siroker is looking at the Web and imagining a future where every page a consumer visits, not just the home page at Amazon or Netflix, has been thoroughly personalized to fit the viewer’s interests or demographic profile. And he thinks he knows how to make it happen.

Optimizely, the company Siroker started with co-founder Pete Koomen, is known today for its automated A/B and multivariate testing tools, which allow companies to show different versions of a web page to different visitors and see which ones work best. But Siroker says some customers are beginning to adapt the technology for general use, rather than just occasional testing. Just as important, he says Optimizely is building up a trove of data about what works and what doesn’t work on the Web—information that future customers will be able to exploit to make their sites stickier and more profitable.

The end result could be a world of fully customized publishing, where every visitor to every site sees a page optimized for them. “A hundred years from now I expect people to look back and be shocked at how impersonal and generic and one-to-many the Web was,” Siroker says. “We have built a very easy-to-use A/B testing tool, but that’s the first step toward a much larger vision of helping businesses show exactly the right thing to the right user at the right time.”

Optimizely employees in their new office at 55 Montgomery Street. Left to right: Julio Bermúdez, Jodie Ellis, Mike Davis, Eric Siroker, Brendan O'Rourke, Dan Siroker (co-founder and CEO), Matt Althauser, Elliot Kim

To pursue that vision, Optimizely just closed a venture funding round involving Battery Ventures, Google Ventures, and Interwest Partners. Siroker isn’t talking about the size of the round—he calls valuations and venture financings “vanity metrics” that are bandied about to “make people feel good at startup cocktail parties but in the end don’t create value.” But he is saying that the 16-employee company is “solidly profitable,” and that revenues have increased by a factor of 12 over the past year. And after just two years on the market, the startup has already signed up more users than Google Website Optimizer, Siroker says.

That’s quite an accomplishment, considering that the search giant’s A/B testing tool is free, whereas Optimizely’s most popular plan costs $79 per month. “Most startups are fearful that Google will get into their market and kill them,” Siroker says. “We did the opposite.”

With huge names like Disney, Starbucks, Crate & Barrel, and eBay on its customer rolls, the startup is now going after what Siroker calls the “800 pound gorilla” in the website optimization market: Adobe Test & Target, the A/B testing tool Adobe acquired when it bought Omniture in 2009 for $1.8 billion. Optimizely is on pace to surpass the Adobe product later this year, Siroker says. “The thing we feel really excited about is the David versus Goliath story—that we can not only take market share from Google but also go after this $1.8 billion company,” he says.

The longer you talk with Siroker, however, the clearer it becomes that what he’s really excited about isn’t A/B testing at all; it’s the idea of helping e-retailers, publishers, and other companies personalize the whole Web. “Companies like Amazon, Pandora, and Netflix are based on this idea of making the Web a one-to-one experience,” he says. “We want to democratize that. A/B testing was a great first product, and we are clearly the fastest growing solution in that space. But now that we have a foot in the door we have a huge opportunity to transform the way the Web works.”

To understand how that’s possible, you need to know something about how Optimizely’s current A/B testing technology functions. A company using Optimizely starts by inserting a short snippet of JavaScript code at the top of the Web page it wants to test. Then the company’s marketing experts can use Optimizely’s editing interface to specify exactly which elements of the page they want to experiment on.

Say you’re with Starbucks and you want to test whether customers are more likely to buy a $50 gift card if you offer them a 10 percent discount on their next purchase. As the gift card page is loading, the JavaScript snippet will run, randomly showing the special offer to half of your visitors (Group A) and the original page to the other half (Group B). By leaving a cookie on each user’s computer, Optimizely can track visitors’ behavior afterward and determine whether the offer affected purchasing behavior.

The key to all this is jQuery, a collection of JavaScript programs that makes it easy to alter the content of a Web page as it’s loading. Optimizely’s system is “very much built on top of jQuery,” which didn’t even exist when Google Website Optimizer or Omniture Test & Target were first written, Siroker says. “We can fundamentally transform a page on the fly, which was not possible even three or four years ago.”

Now imagine that rather than slotting users into Group A or Group B based on random chance, you decide to tell Optimizely to show visitors special content based on what you know about them—say, their location, which is easy to determine from their IP address. All of a sudden, you can test the effectiveness of your discount offer based on geography—which is often a proxy for income and other demographic factors.

By tapping into databases like those owned by Facebook, you can even target messages according to gender and other characteristics. “If somebody is logged into a site using Facebook Connect and the site is using Optimizely, we can use any of the information that Facebook makes available through their API to target a message,” Siroker explains. “We can say ‘Show men this and show women that,’ or ‘Show people who like a certain flavor of ice cream this and show others that.’ That’s something that, even last year, was impossible.”

The Amazons of the world already have the engineering firepower to run their own sites this way. But with cloud-based tools like Optimizely becoming affordable for more and more companies, it’s not hard to imagine a Web that’s essentially one huge, ongoing festival of optimization.

“Our most successful customers start out by asking, what is the best one-size-fits-all solution, which A/B testing is really good at finding out,” says Siroker. “But later they move from one-size-fits-all to segmentation. They ask, what is the best thing for new users? For returning users? For people who have purchased before? And eventually, for each individual user. That is a natural evolution, and that’s why we think this is the future of our company.”

Over time, Siroker hopes, optimization would become not merely dynamic but automatic. Using what it’s learned from tens of thousands of past experiments, in other words, Optimizely could begin to make decisions on its customers’ behalf. “The choices shouldn’t be around ‘This variation is better than this,’ but ‘This is my goal, you figure out the rest,’ using what we know about what worked for other customers,” he says.

All of that may be great for Optimizely, but is it truly good for consumers? In a world of mass personalization on the Web, will there be such a thing as authority or curation or consensus? And if everything is optimized by algorithms, will there any need for people like advertising copywriters or newspaper editors? Perhaps not, to hear Siroker tell it.

“There are some domains that may not be well suited for this, but I think consumers will be the voice at the end of the day,” he says. “My guess is that the New York Times Page One meeting in the future will be a meeting where they say, ‘Optimizely has been running for an hour and it looks like these three stories are the most popular on our site, so those should go on the front page, and we should have this go to the West Coast and that go to the East Coast.’ It’s going to be much more data-driven and objective than having the guy with white hairs on his head decide.” (Sorry, white-haired guy.)

Wesley Chan is a partner at Google Ventures, which just invested in Optimizely despite (or perhaps because of) the fact that the startup is giving Google’s own A/B testing tool a licking. He says it was the analytics background Koomen and Siroker brought to the business—in particular, Siroker’s success helping the Obama campaign raise hundreds of millions of dollars in 2007 and 2008—that attracted Google to the company. But he says it’s the potential for huge financial returns, assuming the technology takes off, that sealed the deal. “I think their immediate focus right now is on building the simplest and best A/B testing tool that anyone can build,” Chan says. “But they will move beyond that. Am I a believer in that vision [of mass personalization]? Absolutely, that is why we are invested.”

“If we can achieve that vision, that ambition to dynamically transform every page on the Web, then we can build a really big business,” Siroker sums up. “I can’t imagine a future where that is not going to happen. It’s just a question of how.”

Wade Roush is Chief Correspondent and Editor At Large at Xconomy. You can subscribe to his Google Group or e-mail him at wroush@xconomy.com. Follow @wroush

By posting a comment, you agree to our terms and conditions.

  • ForumPro

    Fascinating possibilities. I’m pleased there’s a Google alternative anytime I can find one. Call me old fashioned but I’m not a fan of monopolies and consider Google to be the equivalent of the Borg, “Resistance is irrelevant”!