The exact impact of lockdowns and other NPIs on Sars-CoV-2 transmission remain a matter of debate as early models assumed 100% susceptible homogenously transmitting populations, an assumption known to overestimate counterfactual transmission, and since most real epidemiological data are subject to massive confounding variables. Here, we analyse the unique case-controlled epidemiological dataset arising from the selective lockdown of parts of Northern Denmark, but not others, as a consequence of the spread of mink-related mutations in November 2020. Our analysis shows that while infection levels decreased, they did so before lockdown was effective, and infection numbers also decreased in neighbour municipalities without mandates. Direct spill-over to neighbour municipalities or the simultaneous mass testing do not explain this. Instead, control of infection pockets possibly together with voluntary social behaviour was apparently effective before the mandate, explaining why the infection decline occurred before and in both the mandated and non-mandated areas. The data suggest that efficient infection surveillance and voluntary compliance make full lockdowns unnecessary at least in some circumstances.