Jess has a great comparison of faceted and social classification, go read it now. It reminds me of another solution to the metacrap problem: distributed classification. It’s an approach I’ve seen in my last two big taxonomy-driven projects: a financial services firm’s website and a retailer’s CMS.

In the case of the retailer, they have many thousands of products that need classifying on a regular basis. The products are relatively inexpensive commodities that change often and are sold through stores, a print catalog, and online. All the information about the products, including classification, is managed in one big content management system. They could hire a team of librarians to do the classification. Or they could hire less expensive office workers to do it. But they do neither of these, they let the manufacturers classify their own products. When I first realized they were relinquishing control of information that directly touched the customer I was shocked, and now I think it’s brilliant.

This addresses the tragedy of the commons by rethinking the business operations. Instead of looking around their own organization for someone to classify (someone who has no interest in getting it done right, other than being paid to do so), they moved classification outside the organization to those who already have a self-interest in getting it done, the manufacturers whose main focus is selling more products. This is a more scalable solution than hiring a team of librarians. The rules of the system keep the manufacturers from abusing it. And because they have a self-interest in classifying the products they’ll use a pre-existing system, so there’s no need to resort to ad hoc categories.

This worked, meaning 95% of entries were just fine and the other 5% could be easily caught and cleaned up. Classification experts go absolutely bonkers just thinking about this scenario, but this is the scalable way to get the job done. The fact is, the structure of the system was created by non-experts using common sense and the data is added by people using common sense, all without traditional classification training. At the end of the day the retailer tests the findability of items in the catalog with customers and if it works then the classification is correct.

Ultimately the organization of our exploding volumes of information may be less of a classification problem than an organizational design problem. These two clients I’ve worked with pushed the classification outside their organizations and achieved successful, scalable systems.