Where Latkes Live: The Principle of Multiple Classification

In the 15 years I’ve been doing information architecture, eight principles have emerged in my practice. They are not patterns per se — they don’t describe approaches to typical problems — but they do offer a starting point for grappling with difficult IA challenges. These principles deal exclusively with structural issues: they help designers think through the organization of a web site.

I wrote a long article elaborating on the theory behind these principles. In December, I’ll be giving a virtual seminar on the topic.  To accommodate the seminar’s time constraints, I had to eliminate a few principles. That leaves three left over, and I can share one with you in this article without spoiling the seminar.

More than where it lives

Look at any book page on Amazon.com. You’ll see dozens of book classifications, ranging from absolute and prescriptivist (like the Library of Congress Subject Headings) to informal and populist (like tags), explicit (this book belongs to this category) or implicit (this book belongs to a group typically bought together).

Amazon’s book classification uses multiple schemes so you can explore the site in different ways. The principle of multiple classification says:

A web site should offer more than one categorization scheme for browsing content.

As David Weinberger described in Everything is Miscellaneous, ancient philosophers generally believed that the world was inherently categorized. Humanity’s endeavor was to find and label those categories.

Recent developments in cognitive science show that classification is much more complex, and much less absolute. We establish categories for things based on what we already know and what we expect. In short, information architecture requires designing structures sensitive to these observations about human cognition.

Good classification

Early in my career, one of my supervisors established a strategy for our information architecture: “no wrong door.” In short, people should be able to find whatever they need regardless of how they get there. Young, naive Dan didn’t take exception. Today, I would argue that such a strategy could not yield a meaningful classification scheme because users need predictable, repeatable paths, and because different people may think about the content differently.

Epicurious provides a great counter-example to the misguided classification strategy. Recipes may be combined in lots of different ways. To help people find recipes, though, the site’s structure needs both predictability and serendipity, yielding observations like:

  • I will find potato pancakes in the Hanukkah holiday guide, but not the Christmas holiday guide.
  • I can find a menu for my Hanukkah celebration that includes potato pancakes.
  • I can find other potato recipes from potato pancakes.
  • I can learn about the level of effort for this recipe and find similarly-rated recipes.

Today’s classification strategies need more nuance than a simple one-size-fits-all approach. Designing structures with multiple classification schemes demands:

Two kinds of categories

Both bottom-up and top-down approaches, combining an absolute set of categories and categories derived from the content itself. The Hannukah menu is a pre-determined category, but the categories derived from the ingredients are derived from the content itself.

Categories based on rules

Clear communication of belonging vs. generative, such that users understand when content lives in a category and when it’s been thrown in with other content based on a rule. Latkes belong to Jewish cooking, but they’re vegetarian because they contain no animal products.

Linking across categories

Room for lateral linking, such that site editors have the ability to establish meaningful connections between content that isn’t inherent to it. The editors established categories like Cooking with Kids so they can hand-pick recipes that are appropriate.

Classification schemes give users a way to find things, so why confine your site to one scheme? Some users may expect to see certain schemes and others may be surprised by the categories. It’s this weave of expected with unexpected that gives a site’s architecture its appeal.

Users should have confidence they can find what they need. The site’s structure builds the user’s confidence by acknowledging the different ways of classifying the content. By the same token, structures need to delight users by introducing them to ways of looking at the information they had not yet considered.

Learn more at the virtual seminar

Register now for the virtual seminar on the Principles of Information Architecture, where I’ll facilitate a discussion about five other core principles. The seminar will take place on December 15 at 1pm ET.