To design effective instruction, educators need to know "what" design strategies are generally effective and why these strategies work, based on the mechanisms through which they operate. Experimental comparison studies, which compare one instructional design against another, can generate much needed evidence in support of effective design strategies. However, experimental comparison studies are often not equipped to generate evidence regarding the mechanisms through which strategies operate. Therefore, simply conducting experimental comparison studies may not provide educators with "all" the information they need to design more effective instruction. To generate evidence for the "what" and the "why" of design strategies, we advocate for researchers to conduct experimental comparison studies that include mediation or moderation analyses, which can illuminate the mechanisms through which design strategies operate. The purpose of this article is to provide a conceptual overview of mediation and moderation analyses for researchers who conduct experimental comparison studies in instructional design. While these statistical techniques add complexity to study design and analysis, they hold great promise for providing educators with more powerful information upon which to base their instructional design decisions. Using two real-world examples from our own work, we describe the structure of mediation and moderation analyses, emphasizing the need to control for confounding even in the context of experimental studies. We also discuss the importance of using learning theories to help identify mediating or moderating variables to test.