It’s a Messy World Out There
How taxonomy in Information Architecture makes sense of a complex world.
As a child, I cataloged our entire book collection, writing short synopsis and topics on index cards. Little did I know I had entered into the word-rich field of taxonomy, the practice and science of the classification of things or concepts.
Any product or service should want to provide people with meaningful interactions. That means that when we communicate anything, usability through a shared understanding of concepts and categories should be the most important consideration.
And that’s where taxonomy enters.
Any product should be designed, written, and categorized with the end-user in mind because content that isn’t usable isn’t useful.
How about a game?
How would you categorize these?
What was your first thought?
How about this?
This one’s personal:
When I asked my kids, ages 6 and 9, to categorize these, my youngest told me that the T-shirt and the car should go together, because you can be “inside” both of them. While this may not be a conventional categorization, it’s crucial to understand that different groups of people derive different meanings from words, images, and symbols and that it’s up to us to figure out how to communicate the right connotation to the people we want to reach.
As people, we live in a world of symbols, and it’s these signs and symbols that give the world around us meaning and helps us navigate it successfully. Signs are used to create a social reality, and meaning comes from agreement and shared cultural context and experience.
What are some examples of shared cultural context? How about how we get around public spaces.
Just like you use a map at the airport ✈️to find out how to get to the train 🚅
that will take you into the city 🏢, you navigate to the profile section of a website easily when you spot the 👤. But what does 🍔 mean to you? What do the three vertical lines in an app or website tell you? Is that where you will find your account, your orders, how to search? As a symbol, the hamburger menu doesn’t look even close to what it actually hides and as such does not communicate meaning or context.
To help us achieve a shared agreement so we can make useful and usable products, we need to consciously organize the content and the user-flow based on principles that we are able to articulate and that we have collected through evidence gathering.
How do we gather this evidence?
- The Stuff: Data collection — What are we classifying? This is a good place to start. For an apparel company, it would be pretty easy to list off shirts, jackets, dresses, etc, but we may also need to do some research into the corpus of the category.
- The Place: Domain research — How do people who use a similar product interact with it? Are people looking for “trail running shoes” or “Vibram sole Hoka One One.” I.e. is the brand name important or is it a generic search?
- The Lingo: Domain vocabulary — How are people in the industry talking about the product? What’s the jargon? For instance, a manager at a tech company may complain that there is “little cross-functional horizontal alignment,” which for a layperson it can be more easily expressed as “different teams don’t communicate well.”
- The Action: User behavior — Do you have data on existing user-behavior for a product? This can come from interviews with users and stakeholders, customer service calls, help articles accessed by customers, search terms used, and abandoned user-flow or cart. This data helps us identify the frustration points and conduct user research and testing to find better descriptions, categories, or groupings.
Tagging: Another Approach to Taxonomy
The problem with creating hierarchies with single-name categories is that the entities we classify usually only have one category or property attached to it, and as such, they can only belong to one group.
For instance, if we go back to our example of an apparel company, the site may have categories for dresses, blouses, and shoes, but what if a customer is looking for “party outfits” or “everyday basics?” Some dresses and shoes will belong in the party category, others in everyday basics.
The tagging approach to categorization is to attach several properties to the same entities, i.e. to tag a dress as
That way, the dress will show up in results when a user searches for long dresses, green dresses, evening wear and clothes made of silk.
When creating a tagging system, you will want to future-proof it as much as possible, and include properties you may not need at the moment, because it is time-consuming to go back and redo your work later. A good method is to look into the required output first and work your way backward to figure out the scope and need.
Then, break down the elements.
This example of tagging of series and movies on Netflix, show some of the many levels of tags associated with one entity. Some of the tags may never surface for a user, but they are there to build new categories and genres for the future.
- Type (Drama, Romance, Thriller)
- Theme (Family Issues, Revenge, Feel Good)
- Characters (Teenage, Women, Kids, Money Launderer, Inmate)
- Story (Family Issues: Mother-daughter, Father-Daughter, Siblings)
As voice technologies and image analysis become more robust, it will also be possible to extract context and content from for instance closed-captions of a video and images of places, products and people in the video, coupled with an NLP (Natural Language Processing) approach to creating the actual tagging properties.