Current methods of behavioral data collection from mobile devices either require significant involvement from participants to verify the 'ground truth' of the data, or approximations that involve post-experiment comparisons to seed data. In this paper we argue that user involvement can be gracefully reduced by performing more intelligent seed comparisons. We aim to reduce the participant involvement to the 'most interesting' temporal slots, both during the experiment and in post-experiment verification. We carried out a 2 week study with 4 users, consisting of an initial opportunistic gathering of mobile sensor data. Our findings suggest that by using such a method we can significantly reduce user involvement.