In this paper, we propose a way of using optical polarization observations to provide independent constraints and to guide the modelling of the spectral energy distribution (SED) of blazars. This is particularly useful when two-zone models are required to fit the observed SED. As an example, we apply the method to the 2008 multiwavelength campaign of PKS 2155–304, for which the required polarization information was already available. We find that this approach is able to simultaneously describe the SED and the variability of the source, which is otherwise difficult to interpret. More generally, by using polarization data to disentangle different active regions within the source, the method reveals otherwise unseen correlations in the multiwavelength behaviour, which are important for SED modelling.