Not just bought but how much enjoyed

The driver of recommendation engines is not just whether the item was bought, but how much the item was enjoyed. For instance, one can watch Transformers 2, but not enjoy it much. To put it another way, most Americans will say they are against burning the American flag. The question, however, is how far out of their way will they go to stop the flag from burning. They most likely will cross the street to put it out. Will they go to the next town? Will they go across the country? That level is the driver of a recommendation engine.

Customer Want - Not Human Categories

A recommendation engine shows the customer's wants and not a human category. Humans are imperfect in defining a category. For instance, romantic comedy may be a category to market the film. It does not categorize the elements that the viewer will enjoy.

The elements that define a customer's likes are difficult to determine. It could be the resolution of the story, the amount of humor, the type of humor, the style of story telling, the amount of action, the level of dialog.

Instead of displaying the most popular rap albums for the year, the store through a recommendation engine can recommend the rap albums that the customer most likely will purchase. More interestingly, the store can display non-rap albums that a standard search wouldn’t reveal without predictive analytics.