A recently developed statistical technique, often referred to as classification and regression trees (CART), holds great potential for researchers to discover how student-level (and school-level) characteristics interactively affect growth in mathematics achievement. CART is a host of advanced statistical methods that statistically cluster individuals into a number of mutually exclusive and exhaustive groups with markedly different outcome measures, according to the decomposition of interaction effects among explanatory variables. In this study, the author used CART to analyze growth in mathematics achievement during middle and high school. Specifically, the author (a) explored how many distinct groups of students existed in terms of growth in mathematics achievement during middle and high school; (b) measured how variable student growth was in mathematics achievement; (c) described the characteristics of students who grew fast (and those who grew slowly) in mathematics achievement; and (d) identified how student individual-family variables interactively influenced growth in mathematics achievement. The results revealed quite a few findings that are fairly difficult to determine with traditional statistical techniques, such as multiple regression analysis. The ability of CART to identify local interactions (e.g., local gender differences and local age effects) holds great promises for informing education policy and practice. The exploratory scope of this study, thus, has raised issues for potential confirmatory analysis that often lead to the formulation of theories. Therefore, CART can contribute to education research even in the confirmatory sense.