As I write this, it’s a typical Sunday here at Xmarks. The synchronization service continues operating quietly, the servers chugging along syncing browser data for our 2 million users across their 5 million desktops. The day isn’t over yet, but we’re on track to add just under 3000 new accounts today.

Tomorrow, however, will hardly be anything but typical, for tomorrow one of our engineers will start a script that will email each of our users to notify them that we’ll be ceasing operations in around 90 days.

This post attempts to summarize the Xmarks story: how we got to be the most heavily used browser synchronization service in the world and yet still find ourselves pulling the plug.

The Beginning

In early 2006, I built a prototype bookmark synchronizer for Mitch Kapor. We were starting to work together again for the first time in many years, and he wanted me to help him: he was chairman of the Mozilla Foundation but stuck using Safari because there was no way on Firefox for him to keep his bookmarks synchronized across the 5 computers that he regularly used.

The prototype came together quickly and worked well enough for Mitch to suggest that we make it available to others. Curious to see whether it would prove as useful to others as it had to him, Mitch asked a colleague with a widely read blog to write about it, which he did. Hundreds of users showed up to kick the tires and many of them stayed. Some of them mentioned it to their friends. Others blogged about it. Pretty soon there were 5000 users. Of a prototype.

By early 2006, Wikipedia had really started to flourish, a marvel of what could be accomplished by crowdsourcing. It seemed like everyone was reading and digesting Coase’s Penguin to try to understand how an open source community could manage to build something like Wikipedia or Linux. Mitch had the idea that if we traded with our users the personal benefit of bookmark synchronization in return for use of their aggregate data, we might be able to build something useful: a crowdsourced Wikipedia of Websites, or maybe even a spam-free search engine based entirely on what users had bookmarked. We put together a privacy policy that acknowledged the kinds of things we were hoping to do, and set off to firm up operations and infrastructure with the anticipation of growing to hundreds of thousands of users. In October of 2006, we incorporated as Foxmarks, Inc. We had made the transition from pet project to startup.

The Middle

We spent much of 2007 dealing with the growing pains typical of many internet services. We built a team, including front-end and back-end developers, customer support, search, product management, and a VP of Engineering to manage them all. We replaced the off-the-shelf server we had opportunistically used to get started with a custom purpose-built server. As we continued to grow we focused on making the service more reliable and efficient, especially for users with large sets of bookmarks, who were particularly drawn to our offering. We learned a lot about the art and science of synchronization, and poured all of that knowledge into a new client and server which we launched simultaneously and disastrously around Christmas, effectively killing the service for most of our users as we scrambled to understand why the system that we had tested in the lab behaved so much worse in production. Angry users, deprived of the service that they had grown to depend on, demanded that we revert to the previous incarnation, which seemed perfectly adequate to them. We pressed on, and two weeks later the alarms finally stopped ringing.

One of the unseen benefits of the new system was that it enabled us to anonymize, extract, and aggregate bookmark data. So we dove into that and started looking at what products we might be able to deliver powered by the “corpus” of what would soon be 100 million bookmarks. The first thing we built was a search engine. It turned out amazing results, but only for certain types of queries. It was terrible at finding facts. But if you were looking for the websites in a particular category, the results were shockingly complete and entirely spam-free. Looking for the list of all auto manufacturers? Or presidential libraries? Or art supply sites? A casual comparison of our results with those of the major search engines would convince you that we were on to something. We recruited a group of non-technical subjects to do a usability test, and it flopped. Sit people in front of a search box and ask them to test it, and their first query is their own name. #FAIL. It turns out that with the exception of people doing market research, consumers using search are not typically looking for an authoritative list of sites within a category; they’re looking for an answer to a specific question. Undaunted, we tested some variants of the basic search idea, including a version where we inserted our results into the Google search results page. The verdict from users: too complicated.

In mid 2008, the synchronization service was still cranking along, growing at a sustained pace, and that pace ticked up notably with the introduction of Firefox 3. We crossed the million-users mark. Based on our momentum and despite the failure of our early efforts to find gold in the corpus, we secured venture capital funding and recruited James Joaquin as our CEO: “There’s a scalable business in here somewhere,” we told ourselves, and we were determined to find it. James pushed us to find a way to use use our bookmark corpus to enhance web search, an area with a proven Internet business model. We developed several prototpes, and after user testing, we settled on a simple-is-better scheme: we would add information to Google search results showing “bookmark rank” for sites, essentially tallying users’ bookmarks as votes of confidence.

Looking for more growth and a value proposition that could differentiate us from the built-in Firefox Sync that we knew was coming from Mozilla, we invested more heavily in our clients for Internet Explorer and Safari, pushing on the ability to sync seamlessly across these three major browsers. As part of that positioning, we realized we would have to shed the “Fox” naming association we had with Firefox. So at the DEMO conference in March 2009, we rebranded ourselves as Xmarks and introduced the “Smarter Search” feature, as well as a new Xmarks.com website where users could find the top sites across a huge range of topics.

Then we measured and observed user response. The initial behavior was truly encouraging. People using the new versions of our sync clients would occasionally see new “stuff” on their Google search results and click through on links to Xmarks.com. But the novelty quickly wore off and repeat usage after a week dropped off precipitously. We started a series of experiments and systematically arrived at a visual presentation that was more compelling. User engagement improved, but not by the order of magnitude we needed to build a scalable business.

We spent the next year turning over every conceivable rock looking for ways to use the data in our corpus that would prove compelling to our users and revenue-generating for us. Some of these ideas, like SearchTabs, saw the light of day; others never made it out of the lab. Our “SearchBoost“, service was an upsell to advertisers: pay us a fee and we’ll add a mark to your ad when it’s displayed to our users, showing the bookmark rank of your site. Our tests showed that we could boost ad click-through rates by 10%. We built it and it put it front of potential advertisers. Many were interested, but ultimately the feedback was negative: our user base was too small to be worth their time and attention.

The End

By Spring 2010, with money running tight and options fading, we started searching for potential buyers of the company. Over the past three months, we have been remarkably close to striking a deal, only to have the potential buyer get cold feet. We also considered refocusing Xmarks as a freemium sync business, but the prospects there are grim too: with the emergence of competent sync features built in to Mozilla Firefox and Google Chrome, it’s hard to see users paying for a service that they can now get for free. For four years we have offered the synchronization service for no charge, predicated on the hypothesis that a business model would emerge to support the free service. With that investment thesis thwarted, there is no way to pay expenses, primarily salary and hosting costs. Without the resources to keep the service going, we must shut it down. Our plan is to keep the service running for another 90+ days, after which the plug will be pulled.

The past four years have been a wild ride for us: growing something from nothing to substantial scale, providing a simple service that people love because it simplifies their lives. We’ve learned tons along the way, often by making big mistakes. We’re really sorry that this last lesson means that you’ll have to find an alternative to Xmarks, but the alternatives exist and you’ll have no problem finding them. (Start here for specific recommendations.)

I’d like to thank our investors, who stuck with us through uncertain times; my colleagues, who toiled long hours in search of a scalable business; our localizers, who made Xmarks available in 33 languages; and our users, for their unstinting support and willingness to tell us quickly and candidly when we misstepped. You will all be missed.

In the words of Douglas Adams, so long and thanks for all the fish.

Todd Agulnick
Co-Founder and CTO
Xmarks, Inc.