Presenting novel numbers about climate change to people after they estimate those numbers can shift their attitudes and scientific conceptions. Prior research suggests that such science learning can be supported by encouraging learners to make use of given benchmark information, however there are several other numerical estimation skills that may also be relevant in this context. This design-based research project aimed to identify specific mathematical skills that might support postsecondary students' learning from novel scientific data. Concurrently, we also developed an open-source online science learning app. In three design iterations, we conducted 22 think-aloud interviews with undergraduate and graduate students at a Hispanic Serving Institution as they estimated climate change data, before being shown the scientifically accepted value. Productive estimation strategies included: tolerance for error, mental computation skills (rounding and arithmetically manipulating given benchmark values), and integration of prior educational and personal experiences. Two cases are presented, the first illustrates a student who used estimation strategies productively and was tolerant of their inaccuracies, and the second illustrates a student who reacted negatively to feedback on their inaccuracies. Results implicate principles for integrating mathematics and science learning and showcase a learning intervention that embodies those principles.