A method and an apparatus are provided. Distribution information in a first attribute dimension is acquired from a persona data collection of users who use a network service. The persona data collection includes persona data sets corresponding to the users. An evaluation data collection to match the distribution information in the first attribute dimension is extracted from a sample data collection that is collected from sample users. The sample data collection includes sample data sets corresponding to the sample users. Further, a level of accuracy of the persona data collection in a second attribute dimension is determined based on the evaluation data collection in the second attribute dimension.