Where to put the sensors. An Evidence Press audio briefing. Imagine a chain of connected compartments. Material moves both ways between neighbours, enters through selected compartments, and may leak out at specified locations. We can observe some compartments, but the transfer and leak rates are unknown. Can ideal observations distinguish those rates? This candidate paper gives a placement rule for that question, for chains of any length and arbitrary prescribed inputs, outputs and leaks. The rule uses ordered sets and integer counts. It does not need to calculate a large symbolic response Jacobian for each new placement. One subtlety is that different observations can repeat the same information. In one regime, the proof needs a separate augmented count. A seven-compartment example shows why simply adding one to an ordinary count gives the wrong answer. The classification is local and generic. It allows finitely many distant parameter alternatives and excludes exceptional parameter values. It also assumes the full labelled transfer response, known gains and ideal experimental information. It does not guarantee accurate estimates from noisy measurements or a single arbitrary experiment. The package contains a written proof, a replay of saved finite checks, exact minor certificates and review responses. Finite checks and internal model reviews do not certify the complete proof. Researchers can inspect the argument and test the implementation separately. This is an unrefereed candidate dated September twenty-third, twenty twenty-six. The paper and evidence are linked on the release page. This is an AI-generated voice summary, not additional evidence.