How to Manage Successful Crowdsourcing Projects
How to Manage Successful Crowdsourcing Projects
Crowdsourcing is a hot industry buzzword but what does it actually mean? How can it help your business? So many things get called "crowdsourcing" these days, from users voting on content to customer service centers being routed to volunteers. As a result, the term "crowdsourcing" has become tricky to define.
I view crowdsourcing as harnessing the talent with which your business does not have a trusted relationship. These people might be your users or they might be strangers. They might be motivated by money, recognition or the desire to see your product improve. Across the many different crowdsourcing approaches, there are some common themes, both advantages and pitfalls.
Why are businesses relying on the crowd to do important jobs? A simple explanation is cost. In many cases, businesses use incentives other than money to get people to perform jobs that are prohibitively expensive otherwise. For example, Facebook used its users to translate the Website into many of the world's languages. It worked because its users really wanted to experience Facebook in their native language and they were excited about the opportunity to help make it happen.
Another common reason for using the crowd is speed or volume of work. Crowds can often scale up or down as needed, in ways that internal staff cannot. For example, the news aggregator Reddit controls spam by having its users quickly flag inappropriate articles 24/7.
A third key reason companies turn to the crowd is to access many different perspectives. Threadless makes consistently great T-shirts by letting anyone submit a design. Users vote on the best one and the T-shirt is then sold on their Website.
Aside from human perspectives, companies often need access to diverse operating systems and browsers. The Website uTest sources real people to test your Website on different computing platforms around the world.
So, what are the downsides of crowdsourcing? One of the pitfalls that catch businesses by surprise is privacy concerns. Netflix crowdsourced the improvement of its movie recommendation algorithm. The algorithm is critical to its business effectively helping users find new movies-something crucial in order to retain its users. The contest: improve the existing ranking algorithm by 10 percent and win $1 million.
The contest attracted scientists from around the world. The Netflix data was so rich and deep that it promised to yield broader insights about data analysis. Eventually, the scientists formed teams which collaborated to win the prize. Unfortunately, Netflix's attempt to run the contest again was cancelled due to a lawsuit that challenged the release of user preference data. So, crowdsourcing does not always fit the data, and there may be special concerns around private information.
Another issue that comes up is unexpected bias. GalaxyZoo is a Website where amateur astronomers can label pictures of galaxies taken by telescopes. They created the largest database of galaxies ever assembled. The project led to many academic papers and discoveries. However, one finding was perplexing: there were more clockwise spiral galaxies than counterclockwise spiral galaxies. Astronomers wondered if there some galactic Coriolis effect that makes galaxies spiral clockwise. After running a mirrored set of the same images, researchers found that users had a subtle bias towards labeling ambiguous spirals galaxies as clockwise.
More Crowdsourcing Challenges
More crowdsourcing challenges
The way in which the crowdsourcer frames the task will strongly impact the end results. Crowdsourcing allow companies to send work to hundreds of thousands of people. Surprisingly, the language and layout of these jobs dramatically affect the quality and type of responses. Examples of this are Digg and Reddit. Users vote on news articles on both Websites, dictating which content goes to the top. Digg only offers an "up" vote, while Reddit offers both "up" and "down" voting options. In the end, this subtle difference results in two different types of Websites.
In addition, any time you take your projects or data and send it out to a crowd of people, you need to worry about fraud and error. You or your vendor should be sure that you have a system to handle the inevitable subset of people who won't do exactly what you want them to do. They won't either because they're trying to get through the task too quickly, they're simply erring or they're purposefully not doing good work.
All crowds are different, comprised of distinct people. If you rely on your own users, you need critical mass. If you want to use somebody else's crowd, they will often provide recruitment and forecasting in order to match your demand for their services.
Just as cloud computing eliminates the need for businesses to rely on internal computing infrastructure and forecast the need for future computing resources, crowdsourcing eliminates the need to rely on an internal workforce and the need to forecast task volume. For many applications, crowdsourcing dramatically reduces cost and increases scalability, often giving businesses a crucial competitive advantage.
Lukas Biewald is the co-founder and CEO of CrowdFlower. Before co-founding CrowdFlower, Lukas was a senior scientist and manager within the ranking and management team at Powerset, a natural language search technology company later acquired by Microsoft in 2008. Lukas has also led the search relevance team for Yahoo Japan. He holds a Bachelor's degree in Mathematics and a Master's degree in Computer Science, both from Stanford University. Recently, Lukas won the Netexplorateur Award for Give Work, a collaboration with Samasource that brings digital work to refugees worldwide. He can be reached at firstname.lastname@example.org.