Project BoxSand grew from a desire to see if students prepared for a flipped class by watching the pre-lecture videos and to provide free study materials and tools. We built a website with all the best open resources from around the web and built in-house. We study how students engage with these resources and what behaviors correlate with success in the course. We are interested in using Learning Analytics (LA) to close-the-loop with regard to content and pedagogy decisions. We perform Educational Data Mining (EDM) to understand the impacts of those decisions. As we discover more correlations influencing learning, we hope to move into learning more about causation. We are using a mixed methods approach that incorporates both quantitative Deep Learning (DL) and traditional qualitative Physics Education Research (PER) approaches. The BoxSand group is also interested in building new learning tools and testing the efficacy of those tools. One project is a virtual classroom to facilitate Digitally-mediated Team Based Learning (DMTL) in the online environment. Overall the BoxSand group wants to improve physics education by studying what actually works. Learn more by visiting BoxSand.org/about.