Context. In recent years photodesorption rates have been determined in dedicated laboratory experiments for a number of different interstellar ice analogues. These rates are important in order to model non-thermal desorption processes that, for example, affect gas-phase abundances of species and determine the position of photo-induced snow lines in protoplanetary disks. However, different groups using similar experiments have found significant deviating photodesorption values. Aims. Here a new measurement concept is introduced that allows photodesorption rates to be determined following a different experimental approach. The potential of this method is demonstrated using the example of pure CO ice, the solid that gives the most striking discrepancies in the published results. Methods. The new experimental approach uses laser desorption post-ionisation time-of-flight mass spectrometry. It is based on the concept that the physical and geometrical properties of the plume obtained by laser induced desorption of the ice directly depend on the original ice thickness. This allows the ice loss to be determined as function of vacuum ultraviolet (VUV) fluence, which results in a photodesorption rate. The method has the additional advantage that it records all ice species, including photoproducts generated by the VUV irradiation. As a consequence, the method introduced here is also suited to determine the overall photodesorption rate of mixed ices. Results. The photodesorption rate for CO ice at 20 K has been determined as (1.4 ± 0.7) × 10 −3 molecules per incident VUV photon. This result is compared to existing experimental and theoretical values and the astronomical relevance is discussed.