We study the interplay of non-pharmaceutical containment measures, human behavior, and the spread of COVID-19 in Switzerland. First, we collect sub-national data and construct indices that capture the stringency of containment measures at the cantonal level. Second, we use a vector autoregressive model to analyze feedback effects between our variables of interest via structural impulse responses. Our results suggest that increases in the stringency of containment measures lead to a significant reduction in weekly infections as well as debit card transactions, which serve as a proxy for behavioral changes in the population. Furthermore, analyzing different policy measures individually shows that business closures, recommendations to work from home, and restrictions on gatherings have been particularly effective in containing the spread of COVID-19 in Switzerland. Finally, our findings indicate a sizeable voluntary reduction in debit card transactions in response to a positive infection shock.