Our model suggests that distancing measures were working well in British Columbia as of April, 2020, that some relaxation of these measures may have been possible, but that this must be done carefully. More broadly, we estimated a range of effectiveness of physical distancing in other locations (from very strong in New Zealand to weak in California as of May 7). Given the likely low levels of immunity, long-term public health measures will be necessary to control COVID-19. If data were available describing contact patterns and transmissibility by age, along with data describing the impact of specific activities on these contact patterns, then models could be effective tools to determine how to safely relax distancing measures—for example, restarting specific activities such as particular workplaces or schools. Without knowledge of prevalence and transmissibility in children, of the extent of asymptomatic infection, and of the contact patterns that would result from restarting specific activities, models aiming to simulate these activities will have large uncertainties. We therefore suggest that there is an urgent need for longitudinal measurements of population-level prevalence and immunity via viral testing and serological studies, even where prevalence is low. Furthermore, if distancing measures are to be relaxed, it will be crucial to have strong surveillance through widespread testing, contact tracing, and isolation of new cases, as well as strong compliance with potentially shifting public health policy and messaging.