Annual vaccination is widely recommended for influenza and SARS-CoV-2. In this essay, we analyse and question the prevailing policymaking approach to these respiratory virus vaccines, especially in the United States. Every year, licensed influenza vaccines are reformulated to include specific strains expected to dominate in the season ahead. Updated vaccines are rapidly manufactured and approved without further regulatory requirement of clinical data. Novel vaccines (i.e. new products) typically undergo clinical trials, though generally powered for clinically unimportant outcomes (e.g. lab-confirmed infections, regardless of symptomatology or antibody levels). Eventually, the current and future efficacy of influenza and COVID-19 vaccines against hospitalization or death carries considerable uncertainty. The emergence of highly transmissible SARS-CoV-2 variants and waning vaccine-induced immunity led to plummeting vaccine effectiveness, at least against symptomatic infection, and booster doses have since been widely recommended. No further randomized trials were performed for clinically important outcomes for licensed updated boosters. In both cases, annual vaccine effectiveness estimates are generated by observational research, but observational studies are particularly susceptible to confounding and bias. Well-conducted experimental studies, particularly randomized trials, are necessary to address persistent uncertainties about influenza and COVID-19 vaccines. We propose a new research framework which would render results relevant to the current or future respiratory viral seasons. We demonstrate that experimental studies are feasible by adopting a more pragmatic approach and provide strategies on how to do so. When it comes to implementing policies that seriously impact people's lives, require substantial public resources and/or rely on widespread public acceptance, high evidence standards are desirable.