The following is the transcript of my November 5, 2019 Taxonomy Bootcamp DC talk.

Have you ever noticed that superheroes and other crime fighters are better when they work together?

Take the Justice League. Batman, Wonder Woman, Superman, and the rest need to come together to fight the Legion of Doom.

The Avengers cannot defeat the Masters of Evil on their own.

The Ghostbusters. Separately they are a bunch of dismissed scientists. Together they save New York City from paranormal activity and supernatural threats.

The Muppets. Not your traditional crimefighters. But the original Muppet Movie was about finding your tribe and doing great things together. In more recent movies they have also fought master criminals.

I want to introduce two new heroes to you today.

The first is taxonomy: a classification of information into ordered categories.

Taxonomy is powerful. It is obsessively organized and can bring all but the most logical people to their knees.

Taxonomy is sad because, as conceptually perfect as it might be, it might not get implemented or used because the people who need it don’t understand how to use its power. Regular people just want to “tag” things and have magic happen. Those of us who are the creators of taxonomy know that the magic is just an illusion based on a lot of work behind the scenes.

Why is this taxonomy-only approach flawed?

What keeps people from agreeing on them?

I think it’s because it lacks intent and context. Un-or mis-aligned teams don’t know how to apply taxonomy in a way that makes sense. They have a different lens through which they are looking at taxonomy. So they either agree without knowing what they are agreeing to – or never approve a taxonomy.

We need a way to make taxonomy creation approachable and practical.

The other hero is the content model: a representation of the types of content and their relationships.

The content model gets along with others best when presented as a diagram. It can also be a spreadsheet, which the enlightened few love. Like the Philosopher’s Stone that turns base metals into gold or silver, the model captures knowledge and makes it representable.

The model shows structure: the arrangement and organization of parts in an object or system. Like so many other things in our world that have structure—chemical, math, built structures, music—so does information.

Our content model is our first attempt at the structure of our information. The model model has 3 parts:

• Content types: a reusable container for content
• Each content types has attributes: characteristics or qualities of an object
• They are connected by a relationship —in this case an event is held at a venue

Each content type is an entity: a thing or concept that is unique, distinguishable, and self-contained. For example, a performer and a performance.

Each content type is filled up with instances: - unique and specific content, each on an example of the content type. Bono is a performer. Sara Bareilles performing on Nov 19 at The Anthem in Washington, DC is a performance.

The content model provides context and gets to the intent of the classification scheme —2 of the major problems of getting stand-alone taxonomies adopted.

The model is a collection of content types based on what matters most to the context in which the model was made. It should be created at the organization level, not the product or interface level.

It unites all the products and interfaces. It helps to think about the domain first—what subject area an organization operates in. Then the content model is how the organization wants to position itself in the domain.

The domain is the subject area in which an organization operates. For example, the domain of live music could be used by any ticket seller, venue, band/artist, or even an instrument or sound equipment company. For the ticket seller, only some of the domain objects are relevant. For a venue, other domain objects are relevant.

As well, they decide which attributes matter about any given content type. A ticket seller only wants to track three attributes about a venue, whereas the venue itself would want to keep track of many more attributes.

As we zoom in we can see some taxonomies emerging as attributes of certain content types and as content types themselves. And some of the content types share the same vocabulary set.

This makes the taxonomy practical.

It is best to develop a classification approach together. If you are creating your model collaboratively with stakeholders—and you should be—you are now on the cusp of creating a shared language around content structure and taxonomy. The stakeholders can see clearly how classifying or categorizing is more than just a list of words. Not only does the taxonomy itself make sense, but choosing the terms in the vocabulary set will be easier because people already have an idea how it will be applied.

For example, last week I was working with a client to create a content model as part of the creation of a knowledge management program. They had a bunch of content that they called research reports, annual reports, guidelines, best practices, and samples. They thought of them all as separate entities because of how they wanted to present them. Never should a research report and an annual report meet!

But upon closer look at the attributes, we found that they all had the same attributes:


• Title
• Date
• Author
• Description
• Body
• Topic
• File

It was at that point that they realized that we could add just a couple more attributes to describe explicitly what type of thing this was:

• Product Of
• Document type
• Access

This Document content type has a single structure in the back-end, but with a taxonomy applied, it can be delivered in multiple ways to the end-user and in multiple interfaces. If we hadn’t worked together and formed this shared understanding about content types and taxonomies, they would have had five content types instead of one.

The beautiful thing about the model is that it is flat. That means that, unlike a traditional hierarchical taxonomy, it allows you to go sideways, not just up and down. Like a queen in chess, which can move in any direction for as far as it wants to, unlike all the other pieces which are limited to specific directions and number of spaces they can move.

Now you can start imagining more dynamic interactions through the relationships. And that allows people to consider intent. If someone wants to see Bob Dylan in concert, what else do they want?

• More concerts in the same city
• More dates for that performer
• Other concerts of the same genre
• Other concerts at the same venue

The inherent relationships of the model can be used to display the appropriate content to allow intentional linking.

BFFs Solve Problems Together

Together, the model and taxonomies can solve many problems.

One of the most common: Internal search. You can use various taxonomies for filtering and sorting, of course. But putting the structure and taxonomy together, you also affect the results themselves to be the most relevant for the person searching. By weighting certain content types and attributes, you can get more relevant results represented by the most meaningful attributes on the results page (or through a screenless interface like a smart speaker).

They allow you to develop endless interfaces. You can arrange the attributes appropriately on any given template or page based on the priority of user need. The taxonomies make it so you can organize content in a website navigation menu. Take Disney.com for example. You can narrow down which sub-brand in the parent brand you want to explore.

In a product. For example, using the Disneyworld Park app to find out what is happening for the rest of today. What thing, which park, what time?

On a web page. The template for attractions includes decisions about which attributes to show and in what order.

People want what they want when they want it. In other words, not necessarily what journeys you’ve chosen for them. Also, they want a personalized experience. Personalization is not possible without taxonomies and models.

There is a whole talk about personalization later, so I’ll let Jeff and Lindy dig into that. The thing I want to say about it is that it’s difficult. What I hear from CMS vendors that sell personalization as a feature of their product is that the personalization module is why organizations buy it. What I hear from the implementers of those same CMS products is that the personalization module the one used least. Because software only makes personalization possible. Taxonomy and structured content combined with thoughtful humans make it happen. It requires an understanding and judgment of the domain and an audience to make smart decisions about personalization.

Back in the 20th century, the idea of creating a portal or clearinghouse made sense to help people sort through the thousands of web pages that existed. But in the age of Google and Amazon that allow everyone to find anything, those types of websites lose their value. What does provide value is curation—to establish authority and trust as the organization that provides only the most valuable, recent, relevant content for a niche.

We still need humans to do the judgment needed to make sure content is trustworthy, valuable, and relevant for an audience. The choices you make about taxonomies and how you use them to curate content provide clues to your audience about those choices.

Eventually artificial intelligence and machine learning will be able to do this. But we’re not there yet. If you start thinking about how you can do this now with humans and structured content, you’ll be among the first to automate it with confidence at a large scale.

And then there are the programming aspects of sorting through and delivering content from a repository: algorithms, artificial intelligence, machine learning. Sure some super smart people have taught computers to parse the blobs of text that humans have created without giving much thought to entities or their components. But if we really want to teach the computers and allow them to create good results, we have to create our content in pieces, be explicit about the its meaning, and clear in our intent. Taxonomy and its close cousin metadata are essential to that work.

Lessons Learned

If I’ve learned anything in my 20-year career, it’s that the most difficult part of my work is people. Computers do exactly what they are told. People, not so much. So I’ve learned how to get people on board and aligned to produce better content and outcomes.

Start with intent rather than the taxonomy itself. When you go in to someone’s office or a meeting with the goal of (re)making your taxonomy, you instantly lose people. They don’t know what taxonomy is and are entrenched in their own language and view things through their own lenses. They don’t know if the decisions they make about it will give them the results they want.

So stop doing that.

Start with the goal or outcome in mind. Then work backwards to build a meaningful system of classification that helps the people who need it.

Learn how to facilitate the creation of a model and taxonomy. Doing this work is a team sport. An "all play" activity that allows you to make connections with your colleagues or clients by articulating and considering the languages and lenses everyone brings to the table as well as between all your content and systems.

Remember that others don’t know what you know, and they know things that you don’t know. Team work makes the dream work!

Step back from taxonomy geekdom. Isn’t that why we love coming to conferences like this? So that we don’t have to explain what we do and use our common language about organizing and classifying things. It is quite possible that you are the only (or one of the only) person in your organization who cares about spreadsheets, knowledge organization systems, ontology, taxonomy, knowledge graphs, or schemas. Learn to speak the language of the people who need those things to achieve their goals and get their jobs done. It might mean using visual aids and drawing to get your point across—or understand someone else.

Embrace the power of our new heroes: the content model and the taxonomy. One can exist without each other, but together they are a powerful duo.

How will bringing them together in your work change things for you?