#@crowdconf

[ziptask]

@crowdconf in SF. YP.com has their hands in many crowd-cookie-jars.

Daniel Spielman, YP.com

Interesting how YP had so many failures in their pursuit of an elegant crowdsourcing model before finally reaching some level of marginal success in their efforts. They made the most out of on-demand labor by really focusing in on content moderation. Photo moderation specifically. 7,000 photos each week. They seem to have thick skin for this sort of work, which works in their favor for the messy crowdsourcing world.

One point of learning (paraphrased) is they’ve seen specialization among the work categories to be much narrower than before. Some workers have leaned towards more focused vs broad work.

In terms of their structure and process, they’ve done it right. They created their own, internal experts who have served as an internal consultancy to the rest of the business. Compartmentalism works well with crowdsourcing.

[ziptask]

@crowdconf here in San Francisco.

Patrick Phillips, Linked in Andy Boothe, WCG Gordon Anderson, Inside View

The panel seemed to all agree on a few anecdotal ideals: - the crowd is typically “option B” - once you get set up and effective, you can leverage that at scale forever. - if you don’t take the time to create a sophisticated model, don’t expect accurate results. - crowdsourcing is ‘one thing at a time’. As soon as you hear the phrase “and then”, you are doing it wrong. Simple user experience is the only way. - experiments really help. - breaking down (decomposition) of work is a pre-requisite of a successful task design.