Clustering Models VS Preference Models

Let’s say you work for a publishing company deciding upon what New Release Newsletter to send to a particular customer. One simple way would be to just ask your customers as to what genre of books they would prefer to read- Preference Modeling-and the decide accordingly what newsletter to send to whom. Alternatively you can try to predict an individual’s buying propensity for a particular genre and send the newsletters accordingly – Cluster Modeling.
Which of the above should you chose to yield more accurate results?
A survey was conducted with a New York based publisher revealed that deciding which genre newsletter to send to whom based on Clusters had a 2-time better open rate, a 4-time better and 7-time higher click-through-rate, than those based on the consumer’s self-stated preference.
The conclusion: It is equally, if not more, important to observe what people do as compared to what they say.
Read more at: http://www.agilone.com/clustering-beats-preferences