Note: I'm resurrecting this weeks-old posting to make it easier for people involved in a thread on Tagging at the IAI listserv to find it
Observation: Most of the acid test criteria for Web 2.0 applications and online customer experiences include the concept of tagging as a key mechanism for enhancing item-level findability and for making serendipitous connections between items and users within a system, site or application. Customers who choose the same idiosyncratic label/tag to identify the items in a given system may share other interests, beliefs, and preferences in common. Such tagging and finding systems exist as a collaborative filter through which participants may attain efficiencies in finding and re-finding products, information, people etc.
Challenge: What happens if customers start tagging my products with words like "overpriced" or "ugly" or worse? And what’s the incentive for customers to create and assign tags? Once they’ve been incented to participate, how does a website/system collect and aggregate customer use of these tags to enhance item-level findability and serendipitous connections between customers and products?
Especially on a site which provides robust search and browse functionality, how can tagging and folksonometric navigation co-exist with more familiar kinds of taxonometric navigation? In the example above from Amazon you can see that the tags thing is pretty low on the UI priorities list from a page strategy and maybe even a business strategy perspective. Is this short-shrifting of tagging in Amazon's UI due to the scary-ness that free tagging brings to the place where we're transacting with customers? You wouldnt give customers at your bricks-and-mortar store a bunch of sticky notes and a sharpie would you?
Opportunity: In order to side-step the challenges associated with incenting customers to actively assign or vote on tags as part of an online shopping experience, designers are looking to web analytics and customer behavior analysis as an augmentative or even primary mechanism for attaining many of the benefits of tagging. Aggregating data relating to user paths through the site, instances of items being viewed, popularity of items in terms of sales volume … rather than relying on active and willful customer participation via free tagging, all manner of behavior and use on/of the site can be captured and aggregated for use in a back-end "findability engine" from which hyperanalysis of the sum of all online customer activity over time helps shape and adapt site navigation and the surfacing of products.
Concept: Instead of a totally open and user-assigned approach to tagging for online shopping where the site is launched without any tags, waiting for a critical mass of customer activity to populate the tagoverse and unlock the benefits of item-level and fellow-shopper-who’s-like-me types of findability, why not "harvest" or extrapolate adjecteves and descriptors from the existing product catalog on an online shopping site? The products on your site have words associated with them: just turn some of these words into what the Web 2.0 world knows as "Tags."
Customer behavior on the site (viewing products, using search, adding to cart, adding to wish list etc) could be aggregated and analyzed in a manner whereby certain kinds of customer behavior are tantamount to assigning ratings or rankings to products/tags. In such a system, a product that receives more customer page views than its neighboring products will translate into that product’s tags (as derived from its product description) attaining greater rank in the overall system than the tags derived from less popular products.
The tag clouds generated from a site-wide pool of product descriptions would at first constitute a visual snapshot of the vocabulary and language used by the site’s merchants. Once customer behavior analytics and/or customer free tagging is applied to such a system, the tag clouds would reconstitute themselves to become more a reflection of customer interests and community preferences.
Tag cloud navigation for eCommerce is happening right now at Etsy.com. Think of how cool it would be as a merchant to watch those tag clouds evolve over time. Tracking the memes and themes that emerge from your site users' site visits and behaviors could be a tremendously powerful and predictive tool for merchandising and product development.

commenting closed for this article
My name is Dan Klyn, and I'm an information architect.
I work with amazing people at a nonprofit company called Flannel in Grand Rapids, Michigan.
I also teach IA in the library science programs at the University of Michigan and at Wayne State University.