Abstract: TV rating is an important feedback mechanism becauseits results greatly affect the immense profits of TV companies, advertisers and program producers. Program rating with advertising impact has become a key factor in the value of program time. It is desirable to estimate the program ratings fast, economically and accurately. Therefore, how to select the samples for TV rating investigation plays an important role in predicting program ratings. The sampling problem is essentially a bi-objective optimization problem which minimizes the number of samples and minimizes the error of program rating. In this study, we propose an efficient sampling method for selecting samples from a large number of sub-areas. The sampling method utilized the factor analysis based on orthogonal array technique to select a set of the most informative sub-areas with minimal error between the estimated and true program ratings. The samples in a sub-area can be selected randomly. The number of selected sub-areas and the number of samples in each sub-area can be determined adaptively by the decision maker. In this study, the sampling method is applied to Taiwan Cable TV Channels in Taipei and Taiwan. The experiments show that TV rating error of the proposed sampling method is smaller than that of using the same number of sub-areas with the largest TV ratings.