K-12 online education has grown exponentially since its inception in the mid 1990s. Recent evidence of this growth can be found in the number of states with policies towards online K-12 learning (Watson's Keeping Pace), state laws regarding online learning (e.g. Michigan's 2006 requirement that all students receive an online learning experience prior to high school graduation), and the sheer number of students taking K-12 online classes estimated at 1.5 million last year). Unfortunately, there has not been a tremendous amount of research to accompany this exponential growth. There are a number of reasons for this deficiency. First, some virtual schools are so relatively new that data is not prevalent. Other schools that are established are wary of researchers poking and prodding, particularly as many of their offerings are competitive throughout the world. Second, there is a surprising lack of data at many virtual schools. This is due, in part, to the requirements for reporting. If a student is attending a face-to-face school but taking one or two online classes, data reporting requirements fall to the face-to-face school, not the virtual school. Third, and perhaps most important, is that virtual K-12 online education is complex. There are a number of variables that could impact learning, including the teacher, the student, the technology, the course developer, the mentor, etc. Our work in K-12 online education began with a funded study in 2004-2005. However, it took hold in 2005-2006 when we were awarded an evaluation grant from the BellSouth Foundation (later the AT&T Foundation) to study K-12 online schooling. The goal of our grant was to study state-led virtual schools. It became readily apparent that K-12 virtual schools did not necessarily have the infrastructure to collect or analyze data that would inform practice. With this in mind, we developed the Virtual School Clearinghouse. The VSC (http://www.vsclearinghouse.com/) allows virtual schools to download a template of variables about their existing practice. They can then upload the template, filling out as little or as much as they are able, and the system will provide instant analyses of their data. Due to the complexity of K-12 virtual schooling practice, we divided the variables into six broad categories. 1. Students. This collection of variables stored demographic data about the students. It allowed the evaluation strategy to take into account such factors as the number of previous online courses, GPAs, standardized test scores, and individualized learning plans. 2. Teachers. Variables for teachers included such factors as the number of times they had taught online, how long they had taught face-to-face, what qualifications and degrees they had, etc. 3. Schools. A K-12 virtual school's course, say Algebra 1, might actually enroll 25 students from around the state (or the world for that matter). Different host schools provide different support for the students taking online courses. So, in evaluation, it's critical to measure differences with an understanding of the student's context. 4. Course. The course variables relate to the actually course itself. For instance, the course might be Algebra 1, developed by X virtual school, and revised in 2010. 5. Course instance. Where the course was the description of the actual course content and developer, course instance is a given course, during a given time frame, with a given teacher, and a certain set of students. 6. Other. Virtual schools are complex. So, for instance, parents and school mentors arguably play as important if not more important of a role in an online student's education. This final category covers mentors, parents, school administrators, etc. Based on these categories, around 140 variables were created in a template. The idea is that a virtual school can complete as many or as few of the variables as possible; obviously the more a virtual school completes, the more analyses that can be run. Given this framework and the move from BellSouth to AT&T, we now offer free online data solutions to any school. However, we have worked most closely with 6 major state-led virtual schools. Over the course of the last few years, we have collected thousands of points of data. We are also in the process of collaborating with SLOAN-C on publishing a book of findings. The purpose of this presentation is provide highlights from our work. For instance, we have provided evidence that parental belief structures, but not necessarily their direct support, can influence student outcomes. We have provided evidence that teachers who have advanced degrees are more fruitful and stay longer with K-12 virtual schools. We have provided evidence that there is great variation in how schools support or fail to support their students. Finally, we have offered evidence that course authors play a large role in setting the pedagogical stage for their students. Audience members at this presentation will learn from these best practices and will develop strategies for implementation at their own school. They will also learn about the VSC and how to do their own free data analyses.