A method modifies a computer-based interaction based on gaze data. One or more processors collect eye gaze data points to create an eye gaze corpus of information, where the eye gaze data points describe an eye gaze of viewers of a first set of at least one user interface. The processor(s) generate a plurality of clusters of viewers, and determine a target action performance for each of the plurality of clusters. The processor(s) collect, from a device having eye tracking technology, real time eye gaze data from a plurality of current users who are viewing a second set of at least one user interface, and segment the plurality of current users. The processor(s) then modify a computer-based interaction for at least one segment in order to maximize target action performance of the second set of at least one user interface.