Create order out of chaos
Designing content sites is synonym for organising information and making it explorable and searchable. Creating a good taxonomy and gathering the right metadata will help you doing this, but they are never equally important. The question where to focus on depends on the form in which the information is available:

Once an inventory of the current and/or desired content is available, all the information for a site needs to be classified in the form of a taxonomy. The taxonomy is derived not only from the characteristics of the content but also from users’ information needs. Therefore classification in most cases is a collaborative effort: different stakeholders (site managers, authors, readers, potential visitors…) sit together during one or more classification workshops and work out a classification for the available content.
A good classification assumes multiple activities:
The information structure usually is either a hierarchical tree or a matrix. The latter occurs when the information is classified in a faceted way, so it can be searched via multiple facets. In order to define good facets one does not necessarily need to analyse all content, but one has to specify the aspects by means of which each piece of information can be characterised. This principle is also true when specifying metadata for documents.
Classifying information is a subjective activity, and can be impeded by emotional discussions. When an information category is to be given a place e.g. people can have very different ideas about its importance.
The effort of designing a taxonomy mainly depends on the size of the site or intranet, and the diversity of the contained information. Large bodies of content typically require more study time because there is more content to be analysed, and more classification time because the information will often be more diverse and more people will be involved in managing the information.