More importantly, when using the number of responders in the top X decile as the measurement, the distribution within the top X decile and within the rest of the decile can be skewed. For example, if the distribution of the top 4 deciles is 0, 0, 0, and 806, it will give the same score of 806, but that is clearly undesirable. It would be best to use just one number to measure the performance of learning algorithms based on the whole lift table.