CrowdFlower Video

Transcription:

Say hello to Product Manager Pete. As a product manager for a local search service, Pete often buys databases of millions of business listings.

Here's one of Pete's customers. He wants the best steak in town.

Pete wants to ensure that his product serves up fresh and accurate listings and that his customers are happy. But the files he gets, even from the best names in the business, are only 80% accurate. This means that plenty of Pete's customers get really bad results. That accuracy is too low for this hungry customer.

Bad experiences like this have caused some of Pete's customers to leave his site forever.

Frustrated, Pete turns to CrowdFlower. Unlike BPOs and other crowdsourcing systems, CrowdFlower uses proprietary control mechanisms. We screen workers continuously by randomly seeding jobs with errors. Then, we remove workers who fail to fix them. Several workers fix each listing. If they disagree, more workers fix the same listing until the right answer is clear. This technology also means that we've become experts at blocking spammers.

CrowdFlowers accuracy often exceed 95%. These high levels of accuracy means Pete's customers are happy and so is Pete.