Designing Conversational UI with Information Architecture — Part 1

Content modeling to inform your data model.

Welcome to part 1 of a 4-part series I’m writing about conversational UI and information architecture. I’m starting off with content modeling — what is sometimes seen as a humble art.

It’s not sexy. It may actually be the most tedious of IA jobs, but a very good place to start. It is also, I caution, not a simple task.

This post will not be a step-by-step guide on building a content model for your chatbot. It is, however, designed to get you started down that path.

Content modeling is hard because there are many different ways to achieve that same result, but one way will be more elegant than another. You just have to find that way. — Cleve Gibbon

What Is a Content Model?

A content model is not a decision tree. It is not a menu. It is not a hierarchy. It is not a set of editorial guidelines. It is not married to a single web page or device or even user type. Content models are context-independent. This allows us to create scalable, future-proof content.

A content model is the underlying map of related content types that can help create reusable content sourced from a single place.

Start with a Content Audit

How you do your content audit will be determined by what you want to do with your content model. If you’re looking to completely overhaul your organization’s content structure, you may want a comprehensive approach. But if you’re just looking to start somewhere so you can design a better chatbot, you can go smaller.

I am a firm believer that you don’t need to know all your content in order to do a content model. What you do need, though, is a large cross-section of content to help you map it all out.

Build a First Model

If you were designing a chatbot that is giving movie recommendations, you’d want to get a lay of the land. The content audit should help you identify all the attributes of a movie and all the traits of a recommendation. And then finally you would need to identify what qualities/attributes are inter-related and can be used to put together a movie recommendation.

Here is where a lot of information architects will tell you that the level of granularity of your model will depend upon you. But if you’re doing a content model for the first time, I would attempt to use as much of the content you’ve collected from your content audit as possible, and whittle it down from there.


Rework your content model. Add attributes. Remove inter-dependencies. Group them together to form something new. Whatever it is, do something different to see if something else comes out.

Featured CBM: 19 Best UX Practices for Building Chatbots

Why a Content Model Will Help with Your Conversational UI

At this point you may be wondering why this would help with your chatbot.

That movie recommendation content model now has all your inter-dependencies and attributes laid out.

So if a user asks “What other movies have the same director?” your bot can search that attribute [director]. And because you’ve modeled the idea of a recommendation as well, you can not only give one answer, you can give several answers, or something else entirely.

You may decide a recommendation = There are 7 other major motion pictures by David Fincher. Would you like to hear them all or narrow them down by [attribute].

You may decide a recommendation = According to your search history, we believe you’d enjoy The Social Network, also by David Fincher. Have you seen it?

My favorite scene from what I think is a perfect movie — because it’s structured well. Ask me about it!

This type of content structure helps you build out infinitely more scalable answers, whereas a decision tree can usually only work, well, as a tree.

And because a content model is context-independent, you can use the same model and structure for future products.

How Is This Different Than a Data Model?

If you ask developers, it isn’t. A content model can be a mirror image of your data model. But if needed to start somewhere, identifying a content structure should be first, before creating a data model from there.

As with every system and product, there’s bound to be items outside your content model or structure. In a conversational UI, edge-case scenarios may be the majority of them. In a future post, I’ll be talking about hierarchy and giving user options without a menu. Stay tuned!

Suggested Reading

Because this is not a step-by-step process, I recommend the following if you want more details about content audits and modeling:

  1. How to Conduct A Content Audit
  2. 6 steps to product content audit perfection
  3. The Conceptual Content Model
  4. Figuring Out Content Strategy

Read part 2 here, Hierarchy: solving for the lack of navigation.

In February I’m putting together a workshop for World IA Day - Pittsburgh. Come and talk content models with us!

👏👏Clap below to recommend this article to others👏👏