Swivel: Fun for stats geeks and everyone else, with some caveats

I like the idea behind Swivel: Allow anyone to upload any data set and compare it with other data. At this writing, on the service's first day live, there are 13,364 different graphs on the site.

This is fun for stats geeks writ large. The company is opening a fascinating new field of endeavor where the data we collect as nations, states, cities, companies, leagues and individuals will be able to flow together for analytic use. A great pooling of data creates many opportunities for novel analysis and insight.

As a kind of data porn, this will be great fun for everyone other than stats fanatics, too, if they choose to treat Swivel as a curiosity rather than a source of truth. Statistics are, after all,Swivel should consider some "pro" level of service that allows private analysis of data and the ability to publish data and analytics for a fee. the most malleable form of data man has ever invented. You can tell the truth with statistics, but it's also incredibly easy to lie or create spurious "facts" that mask the truth. For instance, this graph on Swivel, which compares violent crime in the United States and wine consumption since 1980 that apparently shows that the more wine consumed the less violent crime happens. But, in fact, wine consumption doesn't have much to do with violent crime, unless you count the incidence of teenage brawls during Strawberry Ripple binges.

Data is fun and profitable. If you can find a meaningful correlation say, between the batting averages of right-handed left fielders during home games in the World Series and global grain prices you might make a killing on the market, but such correlations are usually nonsense. Nevertheless, I believe that in the coming century, so much data will be shared with the world that a kind of occult hobby will emerge, like alchemy used to be, though which folks searched for unlimited wealth and life, in the errata of history.

Swivel is already littered with badly formulated and malformed graphs, like this one, and this one and this one, all of which are literally useless. The challenge for the company will be to surface meaningful analyses using tags and other forms of social collaboration and expert review—both crowd- and authority-sourced insight will be useful and necessary—if Swivel isn't to drown in silly or mangled data. Swivel has tagging, links to other graphs using the same data, and popular graphs. I'd also like to see the inclusion of rankings of graph makers by perceived accuracy and popularity, so that the authors of comparisons are scrutinized.

Now, a couple of outright downsides: The terms of use at the Swivel site grant the company and its affiliates a and assignees the right to use data uploaded by anyone. There's no privacy in Swivelworld. The trade-off is "use our computational cycles, give us rights to the results." It's another situation where user data becomes an asset rather than the basis of a give-and-take relationship.

An Amazon S3 approach to this market, where personal and corporate data is processed for rock-bottom fees but the results remain the property of the data source seems better positioned to create a real market in analytics-based services. At the very least, Swivel should consider some "pro" level of service that allows private analysis of data and the ability to publish data and analytics for a fee, if the author chooses. This would make graphing, like blogging, a potential source of income for anyone with some wits and the diligence needed to build a reputation for accuracy and insight.