Research and theory indicate that the type and quality of instruction, not the delivery mechanism, are what impact learning. Previous reviews have not systematically explored variations in the instructional strategies, approaches, and designs of educational technology programs and also suffered from lax inclusion criteria, which has been known to greatly inflate effect sizes. The present report presents findings from a meta-analysis of educational technology on mathematics outcomes, revealing a moderate effect (+0.12). Statistically significant moderators included study and methodological moderators such as study design, funding source, and sample size. Significant substantive moderators included implementation intensity, implementation context, technology role, and fidelity. Personalization was also a statistically significant moderator.