This paper argues that learning analytics (LA) can make contributions to learning research, in addition to aiming for improvements in learning and teaching practices in a solely applied sense. Indeed, there is potential for transforming learning research if certain approaches developed in LA were used more extensively to study learning. The potential to advance learning research is explored along four dimensions: (1) data quantity, (2) longitudinal data, (3) data from multiple levels, and (4) data from multiple locations. This is followed by a description of how design-based research as practised in the learning sciences can serve as a bridge to LA, and how it would transform LA into a methodology that is also useful for advancing theories and models of learning.