As one of the founders of the original Napster and CloudMark (an anti-spam service), Jordan Ritter, has always been fascinated by the idea of combining the power of the cloud and the potential of human capital. Last April, he met up with Alex Edelstein, a former CloudMark executive, to hammer out their next business plan.
Over a round of beers, the idea for CloudCrowd was born.
Like CrowdFlower and Amazon’s Mechanical Turk, CloudCrowd is trying to bring labor into the cloud, by creating a vast network of workers around the globe. Today, there are roughly 25,000 workers, who have completed more than a million tasks since the site’s launch in October 2009. That represents a small fraction of CrowdFlower’s claim to 500,000 workers, but CEO Edelstein says their force is quickly expanding, with some 2,000 additions each week. Although it took 7 months to reach the one million task mark, Edelstein predicts they’ll reach 2 million jobs in 8 weeks. See video above.
CloudCrowd has two main segments. The bulk of its business is client-driven work: a company submits tasks to CloudCrowd and the service will farm that out to eligible workers. Workers access CloudCrowd through a Facebook app, where they can pick available tasks and arrange payment.The tasks can be as simple as checking the quality of an image or involve the translation of entire web pages. Because pricing is determined by the level of difficulty, the payouts range from one penny to several dollars. Once a task is completed, a different user will check the finished product for an additional fee, creating a level of quality control.
The second part of their business is a new layer of consumer-facing services.
CloudCrowd recently launched EditZen, a simple web site, where consumers can submit pages for editing at $4 a page. The company will soon launch TranslationZen, which will offer translation services at $19.95 a page (far cheaper than some alternatives which are priced at $60 a page).
Aside from their consumer services, CloudCrowd’s cut varies widely depending on the job— ranging from a 20% cut to as high as 80%. The average worker doesn’t get paid handsomely for their task, but it can be a nice source for supplemental income— especially for workers in developing countries where salaries are generally well below America’s mean. In fact, the company now attracts more workers outside of the US, with a heavy concentration in Southeast Asia and Eastern Europe (many from Malaysia, Philippines and Singapore).
The startup, which has yet to reach profitability, is a 15 person operation in San Francisco with modest funding. So far Edelstein is their sole investor, with $1.5 million in the project, but they are talking to major VC firms about a Series A round.
The main challenge for a company like CloudCrowd is keeping pace with its ever-growing workforce. Edelstein admits that feeding the machine with sufficient volume and variety is a difficult endeavor in this early stage. The company has not spent a significant amount of capital on marketing and courting clients, instead, CloudCrowd still relies on word of mouth to attract new business.
“Well it’s true that the number one complaint of our workers is give me more work, I want more availability of jobs. In both of the previous outsourcing, disruptive eras, the era of manufacturing moving to China and developing nations and the era of white collar work moving to India it took years of explaining, educating and showing corporations and all sorts of people who had work how much they could benefit…how consistent the work quality could be,” Edelstein says. “We don’t expect that overnight every single company in the Fortune 500 is going to outsource significant amounts of their work to the internet cloud but we’re sure that it’s going to happen over time.”
The key to their success (and beating rivals like CrowdFlower) lies in the credibility of their workforce and amassing as much information as possible on the skills of their workers. According to Ritter, the system is structured to create dynamic profiles on each user. Every time a user completes a task and that task is reviewed, CloudCrowd will learn how reliable that user is and which skills they excel at. If all that information is organized properly, that could be highly valuable for a client who needs to finish a targeted task in a short time frame.
“We’re building sort of what we call the credibility graph of known reliable people on the internet,” Ritter says. “If there are particular customers who… say, ‘I want a thousand reliable people on the internet to do a particular job for me.’ Well today you really can’t find that. Where would you find 1,000 reliable people on the internet.”