Effectively informing consumers is a big challenge for financial service providers. Triggering involvement in the personal situation of the client is a result of sending relevant information at the right time. While general machine learning techniques are able to accurately predict the behavior of consumers, they tend to lack interpretability. This is a problem since interpretation aims at producing the information a communication department requires to be able to trigger involvement. In this paper we provide a solution for predicting and explaining customer activation as result of a series of events, by means of deep learning and attention models. The proposed solution is applied to data concerning the activity of pension fund participants and compared to standard machine learning techniques on both accuracy and interpretability. We conclude that the attention based model is as accurate as top tier machine learning algorithms in predicting customer activation, while being able to extract the key events in the communication with a single customer. This results in the ability to help understand the needs of customers on a personal level and to construct an individual marketing strategy for each customer.