Research and reform efforts frequently identify evidence as an essential component of science classroom instruction to actively engage students in science practices. Despite this agreement on the primacy of evidence, there is a lack of consensus around what counts as "evidence" in k-12 classrooms (e.g., ages 5-18): scholarship and standards in science education define evidence in a variety of ways including empirical data, analogy, personal experience, and scientific theories. We argue that this disagreement results in a wide range of classroom activities around evidence, including ones that recapitulate traditional science instruction focused on final form science and teachers as disseminators of information. In this paper, we develop design heuristics to inform the design of classroom learning environments that productively use scientific evidence for student sensemaking about the natural world by (i) selecting from the range of information a subset to use as scientific evidence; and (ii) designing classroom activities that support students collaboratively making sense of the natural world. In particular, we argue for three design heuristics that could potentially shift science classroom activities away from traditional "problems of practice" to align more closely with the vision of science as a set of practices including: phenomena-based, transformable, and used dialogically.