The standard process of recruiting online survey respondents is email- or ad hoc survey invitation. With the growing amount of emails and decreasing participation rates this research invitation channels become more and more challenging. This study implemented a new way of recruiting allowing a real-time-behavioral sampling-approach and is far beyond ad hoc website recruiting procedures as it is not only limited to one website and the administrative rights to that website, but is open to any potential web activity that can be tracked. Panelists who installed passive meter software measuring continuously their comprehensive online and mobile internet behavior were recruited via push-notification to participate in an online-survey regarding their in the moment emotional status.

The recruiting process itself happens after a defined event: Therefore a set of 19 websites from 9 different categories (e.g. news, social media, search, media center) were determined: when the panelist reached one of these predefined websites during his natural internet usage the push-notification onto his computer popped up automatically and invited him to the online-survey. The questionnaire contained amongst other scales regarding the emotional status and impulsive shopping behavior.

Results indicate the important role of emotions during the internet usage for both reported and observed data. Three different emotional categories were extracted from the questionnaire data: the disgruntled, the optimists and the burnt-outs. As expected results showed differences in their emotional status regarding various websites. First analysis of the behavioral data showed participants media in the course of the day: therefore each website was categorized by a semi-automatically machine learning method. Afterwards it was possible to represent for each participant its individual digital customer journey for different space of time reducing complexity culminating in nine different content categories.

To sum up: the present studies combined for the first time observed and reported data in a natural real-time-sampling approach. First promising analyses indicate the power of combining both data sources in one research frame.