The Faceted Navigation and Search Revolution

 "The future is faceted." —Peter Morville, President and Founder, Semantic Studios

You've likely heard the buzz by now: faceted navigation truly changes the ground rules for search and KM, and for the many related applications where users need to find information, ranging from subscription content services to directories. It complements search and relevance ranking to fill a big hole in the process of making knowledge reusable. And not only does faceted navigation help end users far more easily find what they're looking for, it also helps content owners efficiently manage content...or, in reality, to make it even feasible to manage content at all.

Below, I'll talk about the distinct properties of Faceted Navigation that create those benefits, and later, explain how classic information science combined with breakthroughs in computer science bring it to life in a form called "guided navigation."

What is faceted navigation? It's a way to browse information, or to refine long lists of search results, along multiple dimensions, aka facets. These are orthogonal lenses through which to view the world. For example, I might search for an expert by facets like name, project, company or date—and more likely, by some combination of those facets, selected in any sequence.

Facets feel instantly comfortable because we naturally think of the world from multiple perspectives. It's like the fable of the five blind men with the elephant, or the tale of Rashomon, where each witness to the same event recounts it from a different angle. There is no universal way to categorize content. Relevance is in the eye of the beholder; the same piece of content might be useful to different people for different reasons, or even to the same person for different reasons depending on their goal.

Busch's Law
So how do facets work their powers? First, we need to state what I'll call Busch's golden law of facets, named for Joseph Busch of Taxonomy Strategies, a past president of the American Society for Information Science and Technology:

Four facets of 10 nodes each have the same discriminatory power as one taxonomy of 10,000 nodes.

That's stunning. That means that with facets, I can describe a collection with 40 nodes (aka subject categories) that would take a taxonomy 10,000 nodes to describe. That's for an idealized case, of course, but the gist of it holds true in the real world. The bottom line is that with facets, we can make do with orders of magnitude fewer nodes than we needed in a taxonomy.

That's because taxonomies are a type of pre-coordinate indexing, meaning that its builder anticipates the compound subjects people can browse along later, like "18th Century French History." In contrast, faceted navigation is based on post-coordinate indexing, meaning that end-users string together their own compound subjects at search time. They do this by combining simple elements from multiple facets, in this example, (Time: 18th Century) + (Country: France) + (Topical Subject: History).

Reducing the number of categories we need by orders of magnitude leads directly to two primary benefits:

1. Faceted navigation helps users more easily find what they're looking for. First, and most simply, it's much easier for users to grok 40 nodes than 10,000, so in practice, they start browsing more. More importantly, faceted navigation offers users many different paths to each item of content—often dozens or hundreds of paths.

For example, before Barnes & Noble.com deployed Endeca, users could browse books through a taxonomy that offered roughly 250,000 paths. Today, with their BookBrowser, Barnes & Noble offers their customers literally billions of paths to browse for books. That difference—between less than a million paths and billions of paths—is the incremental value facets bring.

2. Faceted navigation lets content owners streamline information management processes. Most obviously, if you need orders of magnitude fewer categories or nodes than before, it becomes exponentially simpler to manage them. This leads to cascading benefits. It makes the work of auto-classifiers simpler, since they have fewer buckets to pick from. Also, facets operate independently of each other, which leads to a "schemaless" data model. This makes it simple to combine heterogeneous collections, because you don't need to mesh facets into a single taxonomy, and it makes it simple to add facets incrementally over time.

Facets Pre-date Computers
Librarians have been organizing the world's information since before Tim Berners-Lee was just a gleam in his parents' eyes. Of course, the invention of the Web and intranets has created critical new problems for information scientists to solve, but some universals of knowledge organization persist from the age of bookshelves.