Twitmark V1

Twitmark V1 was the only one that was actually deployed in production.

The story starts mid 2008. My old friend Bastien Quelen and I realized that we couldn’t rely on our old RSS subscription to manage our daily technical watch. Most bloggers we followed had switched to Twitter, eventually turning their blog into a forsaken land, and the people who kept blogging did not share their interesting links the way they used to anymore. We decided to rely on our Twitter timeline to sort and push interesting contents from all over the Web. Twitmark, a Twitter based social bookmarking service was born.

The idea behind Twitmark was dead simple: it fetched every link from our Twitter timeline or favorites, tag them using hashtags and rel='tag', get the title and description from the Web page as well as what other people having bookmarked it had said on Twitter, then push them on our Delicious or Diigo account. We also added some basic recommandation like « people who have bookmarked this or liked this hashtag also liked that content »

For the first time, Bastien and I were launching a public service that was just filling our needs without plotting for world domination.

We had launched the service as a private Alpha. It was ugly, slow as hell, but it did the job and that was not that bad. A few hundred users later and millions URLs later, Bastien dedicated server was crawling on its knees. We had to rethink the whole architecture if we wanted to scale: use a key / value store for the URLs, add caching here and there, move our fetcher model from synchronous to asynchronous… in a word, rewriting everything from scratch.

That’s where we started to ask ourselves the questions that hurt.

Starting from scratch meant renting more machines, or bigger and more expensive machines to profit from emerging virtualization, hiring a designer to build a real identity, and spend more than just our free time on the project.

Twitmark V2

Twitmark V2 was getting rid of the most greedy feature.

We started to apply what we previously learnt when we joined ESSEC Venture incubator. We were not doing that 50 slides business plan with a 3 years revenue forecast to give potential investors a boner but more trying to know whether or not we could break even someday, and even earn some money.

The answer was no.

Too many structural weaknesses

The target niche was way too small: people using Twitter and social bookmarking services who were willing to pay for a service that would help them to gain some time. A small niche is OK when it’s willing to pay. That was a problem.

Another problem was Twitmark very nature. Twitmark was just a bridge between a few service you’re using from time to time, with some cool feature. It was exactly that kind of service that makes you say « it’s cool, I love it, I’ll definitely use it but I won’t pay for it ». Neither Bastien nor myself would have paid for it either.

Twitmark was entirely relying on Twitter, Diigo and Delicious free APIs. Our data source and targets could switch to a paying model or disappear during a week-end and we could not do anything about it. I love the concept of a free – or paying – API, but having my business to rely on data and services provided by another company with neither a SLA nor a guaranty of continuity freaks me out.

The application was simple, too. We had done our V1 in less than a week-end, and it was as easy to clone as a URL shortener service. It had no competitive advantage, and adding one would have meant starting the endless feature race, which was lightyears from what we wanted to do. The feature race to stay competitive is an already lost one, startups that don’t understand it have already failed.

But this was not our bigger threats. Our main competitors were actually the companies we were relying to: Twitter, Delicious and Diigo. Twitter can create its own social bookmarking service anytime. Diigo has built a Twitter connector for years, and since Diigo offers a push to Delicious feature to make the switch from the later easier…

No solid business model

We studied various business models, but none of them were satisfying enough to be worth the effort.

The recurring fee was eliminated before being mentioned. A Flickr Pro account costs 2$ a month, which means we could not ask that much for a service people don’t use directly eve though they rely on it daily. That’s all the « nice but I won’t pay for it » issue.

The first thing we thought about was advertisement. This was before Penguin and Panda, when building for Adsense was still realistic. Since each user has his own public bookmark page, it was possible to create a huge content farm. But this was totally unrealistic. Going Adsense meant bringing an insane amount of trafic to the site, which is expensive with unsure results. The competition was huge, already installed with much more money than we had. And we didn’t want to become yet another crappy scrapping farm.

Plan B included selling the data to third party companies through an API. We were indexing virtually all Twitter, which meant virtually everything new on the Web. Unfortunately, we didn’t have the resources to do it, and Twitter would never had let us do, proof being their API does not allow to do what we did back then anymore. Once again, this was a head of tail strategy, without any control on what was happening. Building a startup is a lot about not controlling everything, but it’s not about not controlling anything.

Plan C was the most serious option. Twitmark could become a B2B application to monitor marketing operations over the social media (2008 baby), by adding more services (starting with Facebook), statistics, tracking, Excel exports and so on… We had the technical architecture to achieve it, everything else was frontend development and UX. But this meant going much further than our initial plan – a small service without too much maintenance – and much more money than we could afford spending.

But these were not the biggest problem. Twitmark was never meant to be a full time business. We just needed it to help us on our daily tech watch, and if we could make money with it, it was through a quick win. What we lacked was enough motivation to get it real.

Perry the Platypus wants you to subscribe now! Even if you don't visit my site on a regular basis, you can get the latest posts delivered to you for free via Email: