Making Crowdsourced Work More Intelligent, If Not More Lucrative

2 min read · 6 years ago


When Matt Bencke led the product team at Getty Images, he says he crowdsourced the massive task of tagging photos so that users could more easily find what they needed in the company’s searchable online database of over 100 million images. But, he says, traditional crowdsourcing outlets such as Amazon’s Mechanical Turk and Crowdflower were “either inexpensive and produced poor results, or incredibly expensive” because it was delivered by high-priced consultants.

High-quality, low-cost crowdsourcing was a resource that he says, “I wanted badly and it just didn’t exist.” So, he left Getty to build it.

Based on “advanced machine learning technology,” Spare5 is the 20-person Seattle startup he now runs and says delivers results “at least 10 times better than prior approaches” to “a few dozen customers” including Getty Images. In a year and a half in business Bencke says, “We’ve completed several million tasks and have several tens of thousands of members in our community.”  

Bencke calls his product “intelligent crowdsourcing” and says Spare5 “takes great care in designing game-like interfaces and pushing tasks to members of a carefully curated community and making sure we’re improving our results.” He says, “We’re helping our customers get insights from a community rather than a traditional crowd.” And, he claims, he’s providing a pastime more fun than Candy Crush and more interesting than checking an Instagram feed for the 50th time in a day. “Get paid for your spare time,” is the app tagline. Bencke says the pay is on the scale of “taco money.”

He acknowledges of most tasks on the platform: “If that were your full time job for eight hours a day your brain might leak out of your ears. But it can be easy and fun to do here and there.” Microtasks could be jobs such as matching clothing in a catalog for retailers, labeling images of pet supplies, describing images of home remodels, or searching online to verify email addresses I a directory.

To be sure, as with traditional crowdsourcing, a living U.S. wage cannot be earned in the intelligent crowd. While Mechanical Turk can pay fractions of pennies for tasks, it took me more than 30 (not-my-idea-of-fun) minutes to earn 75 cents finding attorneys’ email addresses for inclusion in online legal directory of Spare5 customer Avvo. I’d have to go to Mexico to get a taco for that.

Still, Bencke says that 90 percent of the platforms workers, known as “fives,” are in North America, Western Europe, Australia, and Japan as opposed to developing countries. “That’s one way we’re turning traditional crowdsourcing on its head.” And, he adds, “We’ve turned it on its head by saying, ‘This is not a job, it’s a way to use spare time productively.’”