In this study, we present a new insight discovery method for marketing data using our previously proposed approximate inverse model explanation (AIME). Although it has been possible to build predictive models using AI and machine learning for those who are likely to respond to a campaign, it has not been possible to explicitly state why a person responded or did not respond. AIME is a method of deriving explanations using complex AI and machine learning models as black-box models, and constructing approximate inverse operators of black-box models. AIME is a method for deriving an explanation by constructing an approximate inverse operator of a complex AI or machine learning model as a black-box model. In this study, we used AIME to derive the attributes that determine whether a customer responds to the latest targeted campaign using customer data as the target. This method makes it possible to determine what kinds of measures should be applied to which segment of the population responds to the latest campaign, which may help in planning more precise marketing measures. In this study, an experimental system for this method was constructed, and it will be possible to use AI and machine learning to predict the response to campaigns using customer data and to derive the characteristics that contribute to the response/non-response to campaigns using AIME. We also show an example of the output from a representative instance similarity distribution plot that visualizes whether it is easy or difficult to predict whether a user will respond to a campaign based on data distribution using AIME.