Personalization and Information Design

Personalization. It’s everywhere. It’s the big thing. Everyone wants to implement it–and you can’t blame business stakeholders for this viewpoint. Personalization can be a powerful ally in gaining competitive advantage.

What is Personalization?

There are literally hundreds of takes on what personalization is, so I’m going to try to put a box around the subject and contextualize it in the world of information architecture.

Personalization, simply put, is the ability to drive content and services tailored to individuals. A typical implementation of personalization is to match information in a user profile with metadata associated with content. For example, if you join a website and indicate that you have interest in photography and travel, then the website’s personalization system will deliver content and services (and yes, advertisements) that pertain to these two areas.

Here are some other terms that you may have heard within the context of personalization:

• Explicit Personalization: Explicit personalization involves asking site users to fill in forms or otherwise explicitly inform the site what their interests are. To increase success in this kind of implementation, your elicitation techniques had better be brief. For example, instead of asking users to input lots of personal preferences, ask for 1 or 2 very important pieces of data (such as the industry they work in and their salary range) to drive personalization.
• Implicit Personalization: Implicit personalization techniques are more sophisticated then explicit techniques. Implicit personalization involves recording what users do on your site and then displaying more of the same kinds of content. For example, let’s say that you run a financial services website. You track every users’ activities on the site. A particular user logs in and spends a good deal of time researching money market investment vehicles, as well as international stocks. The next time they log in, the front door of the web site will contain links to these kinds of services, as well as featuring advertisements and campaigns directed at that kind of content.
• Segmentation: Segmentation is the art (and science) of slicing and dicing a user population. For example, you may want to segment your audience by salary range, job title, industry, or musical tastes. A lot of sites that claim to be personalized really offer segmentation. For example, just because a site provides content that is targeted to your particular industry and job role does not necessarily make for a personalized site. It is merely a site that knows some pretty broad things about you.
• Collaborative Filtering: Collaborative filtering is the implicit information-gathering of the activities of a community of users that is then used to alter the user experience of those users. For example, on Amazon, users can rate products. These ratings are visible by other users. Also, when you go to a product page, you are shown what other folks who have bought this item also bought. Instantly you are provided with more stuff you can buy. Collaborative filtering is an extremely powerful tool for driving cross- and up-selling.

Are there any problems or drawbacks?

Of course. The number one problem with personalization is that users tend to change their habits fairly quickly, or are wont to try stuff out before settling on a pattern. Take amazon.com for example. I bought Where the Wild Things Are for my nephew a few years ago, and had it delivered. For a while, my recommendations began filling up with other children’s books. The same thing happened when I bought a good friend of mine some cookbooks. I don’t have the slightest interest in cooking, so I had to spend some time quashing those recommendations.

Another issue is privacy. Nicholas Negroponte of MIT once said that a computer ought to know from reading his calendar when he is going to travel to Athens, and drive local news from Athens and other Greece-related content to him as the date approaches. This level of personalization is only possible if we surrender every piece of our lives to the machine. There will indeed come a day when every single step of our lives is recorded in detail, but what web sites and corporations do with that information (which by the way, is more valuable than gold!) is still to be determined.

How does information architecture play in all this?

I know of only one way to drive effective personalization on a website, and that is with a solid framework of metadata that drives a series of personalization rules. What do I mean by that? Let’s take it a piece at a time:

• Metadata Framework: To match a user with content and services, both the user profiles and content must point to the same metadata. You can’t have users of an ancient military website saying they want to see content surrounding “catapults” and other “siege weapons” in their profile when the content metadata uses terms like “stone throwers” and “siege craft”. Nothing would match up. How to solve this issue? You need a controlled vocabulary or thesaurus.
• A controlled vocabulary is a controlled list of preferred terms. It can be a flat list or a hierarchical list.
• A thesaurus is like a controlled vocabulary, but it allows you to add common terms and related terms to the list of preferred terms.

Note: Controlled vocabularies and thesauri are also known as taxonomies and ontologies. Check out Keith Instone’s Personalization paper at Argus for more information.

• Personalization Rules: The set of personalization rules actually leverages and drives the site based on the metadata and what the business wants to do. To return to our ancient military history website example, when a user views a page about catapults, then the website should display:
• Related links to other catapult articles.
• Any events or conferences that deal with catapults or siege weaponry.
• TV show listings for the next month that involve either catapults or ancient weapons.
• Advertisements for special travel tours of Roman battlefields that prominently featured siege weapons.

Notice that these rules not only affect the user experience (i.e., they provide a richer context for the displayed catapult article page) they also manage to drive cross-sell to content and services provided elsewhere on the same site or on partner sites. Driving content to other sites (for say, the special travel tours of Roman battlefields) may drive monetary kickbacks for the referring site.

As you can see, personalization and information architecture go hand-in-hand. A good information architecture foundation (i.e., strong metadata) combined with good business rules can transform a site into a rich user experience.

Where do I start if I want to personalize a site?

Personalizing a site can be accomplished much like any other information architecture engagement.

1. Figure out what the business goals of personalization are. The main goal of any personalization effort should be to build a lifelong relationship of trust between business and customers. But there can be lots of secondary business goals as well: to provide a rich user experience, to drive traffic to partner sites, etc.
2. Figure out who the users are and what they want. Users have problems for which they seek solutions. They come to your site for those solutions. If they are ancient military history enthusiasts, they will probably want to search for content by strategies, tactics, nationalities, and time periods. Ask them what they want to do.
3. Figure out a strategy for creating and delivering the content. This includes creating the controlled vocabulary in order to tag the content. The controlled vocabulary has to make sense to both the content folks tagging it, and the people searching for and reading the content. Be sure to include common or alternate terms as well as preferred terms to increase the likelihood of the content being found.
4. Go into production mode. Create the controlled vocabulary, get the content tagged, spec out the requirements for the web site, have the site coded and tested, and deploy. (I’ve purposefully collapsed a lot of steps here.)
5. Monitor and improve the system. Look at search logs to see what users are looking for and expand your metadata framework as needed (either by adding more preferred terms or providing more alternate terms to existing preferred terms). Do usability feedback sessions with users (or send out email surveys) to find out how (and if) their needs are changing, and improve the site, both in terms of rules and metadata, if needed.

Thomas Myer is the cofounder of Triple Dog Dare Media, an Austin, TX based web consultancy. Triple Dog Dare Media (http://www.tripledogdaremedia.com) builds dynamic web applications, such as shopping carts, ecommerce solutions, and content publishing systems for their customers. Tom loves to talk shop–you can reach him at tom@tripledogdaremedia.com.