Ice nucleation efficiencies, expressed in terms of either the active site density or the ice nucleation rate coefficient, are widely used to estimate ice formation rates in atmospheric models. Most estimates are, however, subject to bias since composition and surface area variation between particles is commonly neglected. This may amount to several orders of magnitude error in active site densities and introduce substantial error in cloud freezing temperatures impacting the accuracy of atmospheric models. Here it is shown that by performing droplet freezing experiments, varying mean particle surface area along with the temperature removes such a bias. The proposed method offers, for the first time, a solution to the long‐standing problem of differentiating the “freezing rate” from the ice nucleation rate, or the apparent and the actual active site density of a material, and will likely improve the estimation of ice crystal formation rates in clouds.