Wait for the numbers has long been my recommendation. Those promoting online K-12 education have made statements about the need for innovation and have suggested that online environments are more responsive to individual needs. I am a fan of the individualization argument specifically when explained as a form of mastery learning (not so much learning styles). You seldom see what I would describe as a mastery system in a traditional school setting.
Studies comparing educational treatments are very difficult. It is nearly impossible to create the right circumstances for a treatment/control study. Without random assignment to conditions, the actual causes of any differences between groups are difficult to identify with precision. You might find quality methodology in small sample studies, but not with large numbers of participants from multiple settings.
Researchers from the Center for Research on Education Outcomes, Mathematica Policy Research, and the Center on Reinventing Public Education have done their best to apply methodological alternatives to a randomized assignment methodology to investigate academic performance of students enrolled in traditional and online charter schools. The results clearly support the traditional approach. While some data suggest specific limitation of online education (e.g., poor supervision) that might be a target for improvement, supporters of online education are going to have to rethink present offerings before contending they provide a superior educational experience.
Center for Research on Education Outcomes (2015). Online charter school study 2015. Available online – http://credo.stanford.edu/pdfs/OnlineCharterStudyFinal2015.pdf
