11. once every three months, an online business-oriented social network company sends out an email inviting its approximately 10 million users to join its paid premium program. the conversion rate to the premium program is about 2%. the ceo of the company decided to change the strategy by offering a better deal to a more targeted audience. the company’s data science team built a predictive model by using historical transactions from the last year. when they applied the model to an out-of-bag dataset (data that was never used for training or validation), they concluded that the model has a lift of 4 at the 10th percentile. however, when the company applied the model to send offers only to a selected audience (top 10% based on the model prediction), the conversion rate turned out to be just 2.5%, instead of the projected 8%. if you are responsible to validate and tune the model, what do you think makes the model less effective?