Sometimes in the race towards automation, we lose sight of the fact that there’s times when, by using people, we can avoid a lot of the errors computers make.  The question to ask then is how do we make human work accurate?  CrowdFlower, has developed one solution to this conundrum, by creating the process and technology to ensure the work is done accurately through a series of questions and statistical modeling.

Recently CrowdFlower launched its Business Listing Verification Solution.  Pitching to enterprises that need “accurate and up-to-date business contact information”, CrowdFlower applies human judgment, across large data sets, to go beyond the inherent limitations of an algorithmic approach. Bear in mind some context with regards CrowdFlower’s area of operation, business listings, statistically over ten percent of American businesses open or close in a given year—thus necessitating a continuous and significant modification of online business listing information.

Created by Lukas Biewald a 28 year old with a stellar record (before starting CrowdFlower, Biewald was a senior scientist at Powerset, and he also led the search relevance team for Yahoo! Japan) CrowdFlower’s Business Listing Verification (BLV) verifies contact information, improves listing attributes such as business name, address, phone number and URL, and supplements relevant data by applying human judgment—rapidly and across large data sets.

It’s a nice combination of human and machine work, leveraging the benefits that both modes can bring. CrowdFlower allows boasts of access to an international workforce of more than 500,000 people from multiple labor channels and more than 150 countries. By leveraging the human intelligence of its workforce, CrowdFlower aims to add a layer of accuracy over traditional data collection techniques—ensuring that the quality of business lists are continually maintained in a cost-effective manner.

It’s also interesting to look at CrowdFlower and contrast it to Jigsaw – the business listing service recently acquired by salesforce – while current fashion is to rely on an automated approach towards data quality – the success of mechanical turk shows that a human component is a worthwhile addition. It’s be interesting to see where the eventual break between automated and human aided processes ends up, CrowdFlower is betting on a mixed approach being most successful.