Using domain models in IA

ByMichele Ide-Smith

Using domain models in IA

Last week the Cambridge Usability Group were lucky enough to have Mike Atherton come talk about information architecture (IA). As a freelance IA, Mike has been working with the BBC for some time to help them tame a huge amount of useful content.

This talk was of particular interest to me. I’ve always had a fascination with how IA’s make complex information structures understandable and easy to navigate. I oversee the management of a large and complex site (a County Council website) which has a lot of varied content. And we need to look at ways to streamline it and focus on user needs.

Mike’s talk was titled ‘Beyond the Polar Bear’. If you have an interest in IA, you might remember the renowned O’Reilly book on IA by Peter Morville and Louis Rosenfeld that Mike was referring to. The main thrust of Mike’s talk was that IA has moved on and UX think needs to be applied all the way through a web project. Information just isn’t neat and we can’t easily categorise information using taxonomies.

Mike also stressed that most user journeys on the web these days start with Google search results or from social media. I know this rings true for the County Council’s website. Google drives around 55% of our visitors.  Another issue which is often overlooked in the world of IA is pages that are mothballed or removed, resulting in a 404 error.

Reflecting on his experiences at the BBC, Mike explained that the BBC has a huge number of different sites, but content exists in silos, so user journeys can be difficult. Also as budgets and staff are reduced the BBC has a need to do more for less. Mike highlighted the issues with focusing too heavily on wireframes – what he described as an obsession with the ‘presentation layer’. Instead he advocated engagement with the fundamentals of content and software development.  The web, as Tim Berners-Lee conceived it, is not about siloed websites. It’s about a wider network, linking content together to enable people to move around following an information scent. Which is where domain models come in.

Domain Modelling

Inspired by Eric Evan’s book ‘Domain Driven Design‘, Mike explained domain modelling as:

“a way of representing the important ‘things’ within a subject, and the relationships between those things.”

He showed a useful example of how different concepts in a subject area link together, to represent the user’s mental model of that domain. An example from the domain of my employer which is easily explainable (local government) is Councillors and Council meetings. I’ve constructed my own simplistic domain model of the key ‘things’ as below (note this domain model doesn’t really express the parent/child relationships between the ‘things’ – but a proper domain model would).

Council domain model

Mike explained how best to establish a domain model.

Firstly as an IA you need to talk to subject experts to discover what the key ‘things’ are in the domain. In my example above the experts would be the officers in Democratic Services. Then you need to talk to users, who often have a simpler language and don’t use jargon. In my example above the users would be citizens and councillors. Mike reflected that it can be hard to get users to articulate what the ‘things’ are in a domain.

At the BBC the terms used to describe the ‘things’ in the domain model are used at all stages of design and development, for example for the database design right through to the CSS classes. This ensures the development team talk the same language as the users. However the developers in this instance do not come up with the data model first – instead the data model is driven by the domain model, because the domain model represents how users understand ‘things’ in the domain.

Search Engine Optimisation (SEO)

Mike suggested only having one concept or ‘thing’ per page, so Google can point to it directly. So in my example that would mean a page for a Councillor, a page for a Committee, a page for a Meeting, and so on.

It’s also important to design URI’s well. Mike felt that IA’s often overlook the importance of thinking about URIs. A good URI should be hackable, memorable and persistent. By ‘hackable’ Mike meant that if you remove part of the URI back to the / it should still work e.g. and should both work.

Mike demonstrated how the BBC had come up with unique, persistent URLs, but recommended the use of ‘vanity URLs’ and a 301 redirect because a unique URL (using a unique code) is not particularly easy to remember.


I am familiar with the issue of having multiple microsites with little to connect the siloed content together. Mike explained how the BBC has resolved this issue by ensuring microsites use the same data/domain model, but with different branding and additional materials, ensuring a smooth user journey between microsites.

Where useful content already exists elsewhere on the web, the BBC use a linked data model link across to 3rd party content. The BBC has taken the view that if Wikipedia has useful content, programme teams may as well ensure Wikipedia data is accurate by updating it themselves.

UX thinking all the way through

Mike likened the need for UX thinking to permeate all aspects of web design and development to Jesse James-Garrett’s ‘Elements of User Experience’ model. In other words, UX design doesn’t start with wireframes – it starts from the underlying domain concepts (the ‘things’). Mike advocated prototyping in HTML/CSS to ensure that you test designs early on with real content.

Search bots are your least able user

We’ve all considered the need to design for the least able user first. But often we don’t consider that your least able user is a machine – a robot! Search bots need clear structure to make sense of your content.

The new IA

In summary, Mike outlined what the new IA should be thinking about and doing. The new IA:

  • Thinks about real-world ‘things’
  • Thinks UX from the ground up
  • Designs for mobile first
  • Designs for search and social media
  • Wrestles data from the web

Here are Mike’s slides:


This content is published under the Attribution-Noncommercial-Share Alike 3.0 Unported license.


About the author

Michele Ide-Smith administrator

5 Comments so far

Nick de VoilPosted on9:07 am - Jul 13, 2011

Very interesting as always, Michele. The funny thing about the domain modelling process that you’ve described is that, under the name of entity modelling or logical data modelling, this is exactly the way that software people have been designing systems for a long time. Because the designer takes it upon her/himself to define what the “real-world things” are, the result can sometimes be systems which don’t support the tasks and activities that users want to undertake, even when stakeholders are involved in the analysis process. The fact is that a “real-world thing” is a very slippery concept and not all people see them the same way. This is why I think that user-centred methods are a helpful counterweight to entity modelling.

However, I do agree that data modelling, or domain modelling, has great potential for aligning what Norman calls the “system image”, the “design model” and the “user’s model”, and this is effectively what Richard Pawson’s “Naked Objects” approach seeks to do. You might be interested in some of the references in this research project proposal from a couple of years ago:

DarranPosted on6:29 pm - Jul 14, 2011

Michelle what an excellent and interesting post! Thank you for sharing this.
I take it that the ‘one idea per page’ concept, does not refer to portal type landing pages?

Michele Ide-SmithPosted on7:46 pm - Jul 14, 2011

Nick – thanks for the comment and link.

I agree it’s very similar to data modelling, but with the distinction that it’s user driven not purely developer or subject expert driven. I know what you mean about Norman’s models. Thanks for the link – will take a look. I haven’t read the book that Mike mentioned (domain driven design) but it’s next on my reading list!

Michele Ide-SmithPosted on8:03 pm - Jul 14, 2011

Darran – the example Mike gives in his presentation of having one page per ‘thing’ is the BBC Programmes site. Let’s say you have the overall concept of a ‘programme’ e.g. EastEnders.

Then you’d have:
A page for an episode –
A page for a character –

I guess domain modelling works well for structured content. But in my view it’s slightly different from the way you design the IA of a website which holds a variety of content where you are trying to optimise user journeys for the main user tasks. That’s where your ‘portal’ style landing pages are useful.

Mike also talked about having an overall concept or ‘canonical thing’, which could then link to variations of that thing. So for example you might think about making lemon meringue pie, but when you search in Google you get lots of results for lemon meringue pie recipes, often linking to the same website. On their recipes site the BBC apparently solved this problem by coming up with an overall page for a common recipe which links to all the relevant recipes for e.g. You can then link to a recipe e.g.

Hope that helps. It’s quite a lot to get my head round!

Mike Atherton on Using Domain Models in IA | The Understanding GroupPosted on12:59 pm - Jul 18, 2011

[…] fans both of Mr. Atherton and especially of incorporating these approaches in the work we do. Do read the full post, and here are Mr. Atherton’s slides: Beyond the Polar Bear View more presentations from […]

Leave a Reply