There are so many large websites which bear the curse of being CMS-driven rather than people-driven. It hangs like an albatross around the neck of visitors.

We know that navigation structures and labels should reflect your audience rather than your organisation. But there’s more to it than that. If you design your website based on content management practices, you’ll condemn your visitors to wander aimlessly through oceans of content rather than make a swift voyage home.

Typically, when you deploy a CMS you do some kind of card sort. You audit your content, group what’s relevant, label the groups and those labels become your putative navigation. If you’ve enough money and sense you’ll test that navigation on some users. This is a pretty straightforward way of coming up with a structure that people understand. But that’s not the same as the navigation that people need. You’ve just prioritised your taxonomy over your visitor’s user journey.

Instead of looking at what you have, ask yourself two questions.

1. Which tasks do the audience most want to achieve on your website?
2. Which tasks do you most want them to achieve (for example, because it saves you money compared to offline channels)?

Those should be two primary drivers for defining navigation. Don’t get caught up in what you have. Focus on what’s needed.

What sort of editorial model do you follow for your web content management? Do you try to get as many as possible hands-on, or do you run everything through a centralised editorial team?

It’s ironic that WCMS which enable you to perform more advanced content management provide tools that you probably won’t want to devolve to part-time editorial teams. Conversely, simpler WCMS are often chosen by by smaller, centralised teams who often feel constrained by the software they use.

Vignette, for example, enables you to assign content to various taxonomies through folders, projects and channels, so that content can be cross-referenced extensively across your site. Put these taxonomies in the hands of people who don’t understand them and you’ll create convoluted user journeys: the exact opposite of your content management objectives.

Alterian’s corporate offering meanwhile — once known as Immediacy — provides pretty basic content management. Most users should be able to get their head around its tools pretty easily. But if you want to create more complex content relationships or have content fragments re-used across your sites, you’re better off with Alterian’s enterprise product, known as Morello. Devolving editorial responsibilities to part-timers who don’t fully understand the consequences of updating content that’s used in lots of places in your websites is decidedly risky, however.

In larger organisations, lots of people will produce content for the web sporadically. These people will change, have variable knowledge of the software and writing style guides, and limited understanding of your website. The last thing they need is a piece of software that allows them to break stuff because they just don’t get it.

So, do you:

1. select a simple WCM for devolved teams to create pages in predefined templates; or
2. select a complex WCM that enables you to perform more advanced content management tasks, but centralise the editorial process.

The more you want to cross-reference and re-use content across your sites, the greater your need for an advanced tool and an expert team to manage it. But if you want to devolve authorship, you’ll need to keep content management tasks and software as simple as possible. Don’t try to industrialise content production by providing everyone with more machinery. For broader participation you need to provide hand tools. Leave the combined harvester in the hands of experts.

I think that byline was a bit facile, but the article does prompt a serious debate about where taxonomy and more particularly expert taxonomists are heading. When organisations think web 2.0, they think wikis, user-generated content and tagging. They don’t think about well-organised content. Web 2.0 implies the death of expert taxonomy rather than the death of taxonomy itself.

People just don’t want taxonomists; they think they can organise content for themselves. While this may be true, it’s unfortunately also true that they just can’t organise content for anyone else. This creates a particular problem for systems which are dependent on finding very specific information: intranets, for example.

If you need your audience to be able to retrieve information reliably, don’t look to “audience development officers“: recognise that taxonomists do have an expertise that you’ll find useful. However, not all information needs to be structured, so feel free to challenge any taxonomist who tells you otherwise.

Firstly, by providing a simple search box. After all, this is the way most people find new information through Google. Secondly, by using a search tool to push similar content to users; for example the right-hand column of this page provides links to pages that are classified under different themes. But before you view search as some kind of panacea for all your information architecture woes, let’s pause to reconsider these two methods.

In the first case, how do you know that the search results presented by Google are the most relevant pages to your query? Google has no real benchmark. Then weigh up how much effort people put into ensuring their content is optimised to appear at the top of the search results and then ask yourself what you’d have to do to ensure that relevant content for every search a user undertakes.

In the second case, consider that actually I’ve already (very loosely) made decisions about navigation by tagging every post. This is almost the same effort as organising the content into folder structures as you would in a CMS. In fact, for sites with lots of content it can be more difficult to tag all the content than to drop it into a folder structure; the folders provide a more complete classification metaphor that’s easier for people with less expert knowledge to implement.
So how do you decide when search is better for navigation than in a CMS? Here are my suggested criteria:

1. You have the money.
   Implementing faceted search technologies can be significantly more expensive than using standard content management system functionality. Day rates for leading product professional services are often relatively high, there are licence costs and there’s an additional cost of integration, particularly if you need to tie a security model into the search tool.
2. You have few content types.
   But you have lots of content. Structured navigation from search works well where you have similar kinds of content, with similar structures against which the search engine can execute straightforward queries. A product catalogue is an obvious example. The tool can filter on price, format, location, etc. which have definitive and distinct values.
3. Your content is distinct.
   Categories need to be unique; semantic tools aren’t really advanced enough yet to tell you that apple is a fruit not an iPhone when displayed with orange unless orange is the network provider. Moreover, your pages need to be called something readily identifiable. If you have ten pages called “Help” or “Contact Us”, how will the search tool know which is the relevant resource for the site visitor?
4. Linking is obvious.
   If you use search to provide your navigation, you relinquish editorial control, so it must be clear why pages in the navigation are related. On a medical site, for example, you might link to other conditions treated with the same drug. However, as soon as you’re trying to tell people something you can get in trouble if you automate. An example I often use is a page about health advice for eggs: should this link to information about required protein intake (i.e. eggs are good for you) or about cholesterol (i.e. eggs are bad for you)? Or should you exercise some editorial discretion and explain about balanced diets? There are few search engines that could perform the navigation required to achieve the latter example.

Practically, it’s generally simpler to use CMS to navigate, with a search option to help people who are stuck. The advantage of CMS-driven navigation is that the editorial control you can exercise should help you to push visitors to your site along a route you want them to take. However, if you’re happy to let your intrepid visitors explore your content, and you’ve nothing in particular you want to promote to them, then search engines can be a viable means to provide navigation.

My final analogy is that CMS-driven navigation is like a library, while search-driven navigation is more like a bookstore. In a library you’ve preplanned how visitors can find specific information. In a bookstore you’re encouraging them to browse, but they may never find what they came in for.

1. Identify key user types (personas)
2. Identify key tasks they need to undertake (user journeys)
3. Develop navigation to enable journeys (site maps)
4. Develop user interface that will enable users to complete journeys (wire-frames)

Main advantages of doing things this way:

• You’re not trying to fit in existing content unless it’s actually useful to your users.
• You can identify content that’s missing easily.

There are more useful IA definitions at iaonesheeters.com

To be able to tell people that FAQ content is available on your site.

Who do FAQs help?

Only the person who needs to claim that the content is on the site.

Why don’t FAQs help visitors to your site?

Because visitors to your site don’t care whether their question has been asked by anyone else or not. All they want to do is accomplish a task.

But my FAQs are representative of what all visitors are asking, so don’t they bring value?

Think of it another way. Your site is providing a user manual for your services. By providing FAQs you’re offering them that user manual without a contents page. Is that helpful?

But isn’t it helpful to provide what most people want first?

Of course, but if I don’t want the most obvious answers, I’ll give up and go to another provider: either through a competitor website or I’ll pick up the telephone and harass your call centre. That probably defeats the purpose of your website.

Why is browsing FAQs a flawed argument?

Let’s say you’re describing your arrangements for dealing with business partners. If I want to become a partner, I should follow links to Partners > Becoming our partner. Why would I trawl through a long list of questions in order to find the right one. I want a simple and obvious path to the information I need.

Why can’t visitors just search the FAQs?

They’ll probably do this if you offer no other way to the information. But the success of their searches will depend on how well and how consistently you classify the questions and on having a really good search engine that will pull out relevant information.

Isn’t it easy to manage FAQs?

It might be easy to add an FAQ, but it’s extremely difficult to manage them. You’ll need to check that similar content doesn’t already exist, just framed with a different question. You’ll need to check that you’re removing questions that are obsolete, or simply not being asked any more. And you’ll need to ensure that questions are presented in the right order according to your site visitors’ behaviour.

So is there anything good about FAQs?

Not in their standard format of unsorted lists of questions and short answers.

So what should I do?

Review what people are contacting you about over other channels. This information should probably be on your website. Does it exist already? If it exists, is it being adequately promoted? Undertake some usability studies challenging people to find the information. What do you learn about your site navigation from this? If the content is prominent, is it well-written? Are people finding the content and misunderstanding it?

Isn’t this post just stating the obvious?

Yes, but it’s amazing how many people think their website needs FAQs but they never ask themselves why.

Taxonomy is about classification. It describes ways of naming, arranging and ordering things within a system. Those things may be books in a library, or plants and animals in biology.

Taxonomies are usually hierarchical and based on restricted terms. In biology, for example, there are a limited number of kingdoms, phyla, and classes. In the Dewey Decimal System, these restricted terms are based on numbers: Phonology (414) is an element of Language (400). These relationships are often described as ontological.

Some things can belong in more than one place in certain classification systems. For example, a dish on a menu may be available as both a starter and a main course. These ways of looking at the item are often termed “facets”; so a Greek salad is both a type of dish (salad) and a course (starter).

The “thing” in question may also have one or more synonyms which form part of a taxonomy. There is some overlap here with the functions of a thesaurus.

A recent innovation brought to the fore by websites like del.icio.us and Flickr is the concept of “folksonomy”. This allows participants to “tag” any page, image, or document according to their own vocabulary. As more people tag pages, these become related by shared terms, building up the classification system. This taxonomy is unrestricted: what one person calls design, another may call art, creativity, or even Photoshop.

How are taxonomy and content management related?
Content management typically addresses a number of key areas:

• Production: providing the tools to enable people to create content.
• Authorisation: ensuring that only relevant people are able to view and amend content.
• Workflow: delivering content to people within the system once it passes through certain stage gates.
• Storage and retrieval: providing a mechanism to store, find and re-use content held in the system.

Content management systems (CMS) can make use of taxonomy for authorisation and workflow, but they are dependent on taxonomy for storage and retrieval. Each page, document, or other asset held in the CMS is stored according to a predefined classification method.

There are different metaphors for this method. The taxonomy may be represented as types of document (e.g. contract), as departments with their own document silos (e.g. legal, marketing, human resources), as folders that represent some other business function (e.g. project start-up, initiation, execution, closure), or flagged with a value from a predefined list. Most CMS will use a combination of these classification systems so that content can be retrieved more easily.

Taxonomy vs. metadata

Taxonomy is often applied as metadata: that is data about data. Office documents have metadata assigned to them such as author and revision date. (To see this in your Office application choose File >> Properties and click on the summary tab.) Web pages also have metadata. There are a number of ways to view this, but the simplest is to choose View >> Source in your browser menu. Close to the top of the page you’ll see some HTML tags beginning meta name=”". The name attribute is the type of metadata they describe (description, copyright, keywords, author, etc.) while the content attribute holds the metadata itself. Metadata is often assigned from pre-designated classifications, so is an important part of any taxonomy project.

Where can I see taxonomy on the web?

In its simplest form, taxonomy on the web is represented as website navigation. Many of the techniques applied to develop taxonomy are used to develop more user-friendly websites. This includes activities like card sorting.

More complex taxonomies can be found at dmoz.org (a directory of websites) or, for medical research, GoPubMed.

It’s down to psychology, not technology.

Let’s refer to the Mirror Stage, a psychoanalytical concept developed by Jacques Lacan during the 1940s. The concept describes how infants imagine themselves to be at one with a mother who satisfies their every need. When a child cries, its mother will feed it, change it, put it to bed, or comfort it. When a child reaches between six and eighteen months old, it starts to realise that its identity is separate to its mother’s. It recognises itself in a mirror, has to learn to feed itself and will be told off by its father. In short, it enters a symbolic order where it now has to conform to social constraints in order to get what it wants. The early imaginary state provides gratification without context, while the symbolic order provides context but imposes boundaries.

Which psychological order do your contributors belong to? Do you want or provide an environment for them to express themselves freely, or do you need to contextualise them and the content they produce?

Collaborative tools assume a shared identity. Just as an infant considers its mother to be an extension of itself that responds to its every whim, users look to collaborative software as a personal tool that instantly fulfils their need for self-expression. In this imaginary order, contributors “write out their question in their blog and look for their community to respond and help them“. Compare this to a content management system, where you have both context and boundaries: contributors recognise that their content can only be published if it meets predetermined social criteria.

Some examples:

• Folskonomy vs. Taxonomy: The most obvious difference between imaginary and symbolic orders in content classification. In folksonomy, users enter terms that help them understand their content and they imagine that other users understand these terms. In taxonomy, these terms are given a context and only predefined terms can be used according to a preordained structure.
• Intranets: Is your intranet an environment for generating knowledge or enshrining it? If your staff use it to discover what’s going on across multiple locations and projects, they assume that content is representative of the work they do. If the intranet holds authoritative information that employees want to refer to (for example, HR policy), you need a tool that confirms their place in the organisation and that reasserts social context.
• Web 2.0 vs. Web 1.0 sites: People who use social networking sites subconsciously assume that what is valid for them is valid for others: that their tags make sense, that their ratings (of YouTube videos for example) are relevant, that people will follow their myspace page. These assumptions may well be right, but context is limited to these assumptions. If I put a photo of Sophie onto Flickr and tag it accordingly, this tells me that there are other photos of people called Sophie on the site, but doesn’t tell me that it’s Sophie Marceau and I’m interested in pictures of French film actresses. It’s not clear of course that this is the information people are looking for, but a content managed system would presume this in its design. So if you go to a report on a football match on the BBC news website, it will provide links to more news about each club involved, league tables, fixtures, weather forecasts for that area and so on. The contributor doesn’t elect to have all this correlated information: the CMS provides the context automatically and imposes an authoritative order.

Of course, collaborative environments aren’t completely without context: any user who logs into the system has a distinct identity within the organisation. But the mirror stage in content management comes when you start to impose structure and workflow. If you need your contributors to put content in a specific place for easier retrieval, or to have their contributions approved before they’re viewed by a wider audience, then you’re imposing a symbolic order.

So when choosing your approach, ask yourself are you a mother or father to your users? Are you coaxing them, encouraging them to express themselves freely, or are you imposing a paternalistic authority?

If your organisation is essentially the same thing as your contributors, then an unstructured wiki is a viable option. This covers social networking sites, or collaborative research intranets. But if your organisation represents something more than the people it comprises, in line with a Gestalt psychology, then you need a content management system that enforces a shared identity rather than assumes it.