Abstract When planting flowers for pollinator conservation, determining what flowers to plant is challenging because flower establishment can be time‐consuming and resource‐intensive. To alleviate this challenge, researchers have proposed methods to mathematically determine from plant–pollinator interaction data which flower species pollinators prefer, which can be defined as the likelihood that a flower species will be chosen by pollinators when offered on an equal basis with other flower species. We compared the flower lists produced by five sensible, peer‐reviewed preference metrics calculated from the same dataset and examined how each metric controls for flower abundance and relates to number of pollinator visits. We found little correlation between the ranked flower lists returned by each preference metric and that the metrics varied in the extent to which they controlled for abundance and provided different information than number of visits. The discordance among calculated flower preference lists is partially due to the different way each metric controls for abundance and suggests that these preference metrics need to be empirically tested and that more research is needed into the factors that impact pollinator floral preference. We discourage the use of three preference metrics (confidence interval, resource use and mass action hypothesis metrics), caution against the use of one (centrality metric) and recommend the use of the preference index metric due to its insensitivity to insufficient sampling, ease of use and the fact that it is not correlated with the number of pollinator visits.