We use a unique combination of register and survey data from Amsterdam to investigate the performance of school assignment mechanisms in prac-tice. We find that deferred acceptance (DA) results in higher mean welfare than the adaptive Boston mechanism. This is due to students making strategic mistakes. The welfare gain of a switch from actual Boston to DA is over 90% of the welfare difference between actual Boston and optimal (proxy) Boston. Disadvantaged and lower-ability students would benefit most from such a switch.