Reported sensor performance is a context-bound claim, not a property of a device, yet headline metrics are routinely reused as decision-ready evidence. This hands-on course gives participants a four-layer method for reading any performance claim: who ran the evaluation, what data product was evaluated, under what conditions, and what the metrics do and do not convey. Working in groups with real anonymised evaluations and use cases, participants produce a one-page evidence map for a claim of their own and leave able to identify what is missing and whether it matters for their application. Here we will use low-cost air pollution sensors as an example, but will be relevant to anyone across disciplines procuring, deploying, or citing evaluations of environmental measurements. This session is organised by the Allin-Wayra activity of the International Global Atmospheric Chemistry project.