Learning progressions (LPs) are theoretical models of how learners develop expertise in a domain over extended periods of time. Recent policy reports have touted LPs as a promising approach to aligning standards, curriculum, and assessment. However, the scholarship on LPs is relatively sparse, and the jury is still out on the theoretical and practical value of this approach. To realize any potential of LPs researchers need to systematically refine these hypothetical models in real-world contexts. Such refinement efforts are challenging, as they require the coordination of messy empirical data with often underspecified theoretical models. Many of the current reports involving the empirical refinement and validation of LPs do not sufficiently explicate the process of how one goes about making modifications to the LP based on empirical data. In this article we present heuristics for facilitating the coordination of data and LP models. Using an illustrative example of a genetics LP and data from a 2-year longitudinal study of this LP, we demonstrate the use of these heuristics to refine the hypothesized levels of the LP. We also discuss the process we used to identify contingencies (relationships) between the constructs of this LP. We conclude with a discussion of implications of the refinement process for the alignment of curriculum, instruction, and assessment. (Contains 4 tables and 2 figures.)