The paper proposes a methodology for the development of a marketing decision support system using Big Data technology and data mining techniques. The approach was inspired by the CRISP-DM methodology, which is not oriented towards Big Data projects. Therefore, we have modified this methodology with respect to the purpose and technological requirements of the project. The proposed methodology was tested during development of RTOM (Real Time Omnichannel Marketing) project. Project tasks focus on the analysis and exploration of large and heterogeneous data sets. The paper presents the phases of the project implementation according to the extended CRISP-DM methodology, taking into account the specifics of the analysis and exploration processes of large real-time marketing databases. Examples of project steps are also provided to illustrate the approach.