I’ve seen a few projects now, mostly big ones, where we’re trying to create generic metadata behind a website, the idea being you can then use that information for other applications (an example of an application being a portal that uses a person’s customized preferences to filter a set of documents using metadata associated with the documents). But there always comes a point when we have to put a stake in the sand and accommodate the application at hand. Eventually the metadata ends up being customized to work for specific applications and not just any old application. So my current thinking is that we should ignore all but the applications at hand. Metadata can be massaged and schemes can be morphed later to accommodate other applications, or at least that’s the assumption I’m going on for now (please tell me if you know better).

So, rather than start with a metadata scheme, it makes more sense to work backwards starting from the user interface…

• do all your up front user-research
• determine what information will populate the user interface
• determine what kinds of metadata will be necessary to pair that information with those users (or uses)
• given the metadata needed, devise the metadata scheme to organize it all