Personalization is the idea of tailoring an experience to a specific individual. This is particularly prevalent in e-Commerce. There are a few ways to achieve this.
The ATG e-Commerce platform uses ‘a scenario engine’ to provide personalized content. The engine uses a rules-based system, either based on user attributes or user behaviour. These can be simple (“the user has set her preference to be movies, so we will show her movie-related content”) or incredibly complicated (“in the past week, the user has clicked on at least three movies and at least four television shows, so we will show her randomly served up movie or tv-related content”).
Such rules-based systems are powerful, but they also necessitate a fair amount of planning and development. ATG actually recommends a specific job role dedicated to writing and configuring such rules. So what if you’re not entirely sure what the specific rules you need are?
An alternative that is gaining ground is based on Bayesian probabilistic theory. Don’t worry, there’s no need to actually know the name of it 🙂 Bayesian theory “us[es] the knowledge of prior events to predict future events.” This is the basis of the Cleverset product recommendation offering (Cleverset has now been purchased by ATG to round out its offering). Bayesian theory is (perhaps unsurprisingly) well-discussed within academic circles. A few example papers include: “Online Recommendation Based on Customer Shopping Model in E-Commerce”, or “Website Morphing”.
Many people have an amazon.com recommendation story: they purchased a baby item once as a gift, and continue to have baby items recommended to them on each visit. A rules-based system such as ATG allows you to set thresholds (you must buy at least three baby items – perhaps even on three different visits – to have baby items recommended). A system that uses Bayesian reasoning basically modifies itself based on your behaviour to predict the likelihood of your buying another baby item.
The above example is based on targeting content to a user, based on activity. When I first came across an article referencing Website Morphing, I was quite intrigued. Rather that focusing on content, Bayesian theory is applied to the cognitive style of the visitor. Whereas our goal for a particular site visit may change, we all have preferred learning styles. How I choose to gather information on a site can dynamically modify how it is presented to me (I am adverse to video tutorials, so I will enjoy a site more if that information is provided differently). According to the MIT Sloan professors that are doing this work, “morphing websites can increase sales 20 percent .” I have been in email contact with the author of the paper, and he states that they have continued to try out their system in limited markets, and continue to see the same positive results.
Personalization is one step in providing an optimized user experience. When looking to incorporate personalization into an experience, consideration must be made both as to the specific approach as well as the nature of the personalization (is it content or presentation that is tailored).
 What is Bayesian Logic? retrieved July 7, 2008 from http://whatis.techtarget.com/definition/0,,sid9_gci548993,00.html
 MIT Sloan Professors Find That Morphing Websites Can Increase Sales 20 Percent retrieved July 7, 2008 from http://rismedia.com/wp/2008-05-22/mit-sloan-professors-find-that-morphing-websites-can-increase-sales-20-percent/