The Basics of Web 2.0 Personalization

I gave a short talk internally the other day on personalization, which is one of the important concepts many of our clients ask about. It was very basic, essentially roughing out a framework for thinking about personalization from an information architecture perspective, and pointing out where some of the Web 2.0 twists are. Here's the cliff notes version:

 

Web 2.0 is made of two important things: people, and content objects. People are represented by identities. Most of us have many. Identities have information attached to them. This picture is of a person. The person has attributes.

 

When a person comes to your web site, you often know a few things about them, like their referring URL, their browser, etc. This isn't much. An early goal for conversion on Web 2.0 sites is account creation. This picture is of a person with a brand-new account. You know their account information, but not much else. The person has many attributes, but you don't know what they are.

 

This is your web site. It's made of content objects, and the content objects have attributes. In Web 2.0, you know what some of the attributes are. Other attributes, like user-generated tags, might not yet be knowable.

 

One of the most basic kinds of personalization is access control. Based on the person's account information, they can access some content objects and not others.

 

As a person uses your web site, you learn more about their attributes. Sometimes they tell you ("subscribe to newsletter") and other times you learn about people by knowing what they do ("I bought Scoble's book about blogging").

 

Different content objects have different attributes. Some, like content type (photo, video, downloadable document, product info) are structured, and others, like title, are unstructured. Some, like tags on a Flickr photo, are added by users.

 

Basic content personalization is a matter of matching up the attributes of a person with the attributes of a content object. You know the person likes basketball because they bought a Duke basketball jersey, so you show them a content object that has to do with March Madness.

 

There is some complexity in the details, I have to say, including reconciling attributes that don't have 1:1 relationships, such as when a person's location is "northwest" and your content is tagged by state.

Anil Batra has taught me a lot recently about using complex behavioral targeting mechanisms to personalize content--for example, if a person on a newspaper site looks at the financial section and the automotive section, you can target them with an ad for a Lexus. New services are emerging that enable you to track behavior across multiple web sites and use that data for personalization. Combine that with advances in identity management, and you're in the brave new world.

And some of the details are fundamentally different on Web 2.0. Content objects are created and assigned attributes by users of your site. User-generated content and folksonomy change the game for personalization. I believe many, many corporate web sites are going to learn these challenges quick--and the folks who get it right first are going to stand to gain a lot.