Taking advantage of the unprecedented combination of sensitivity and angular resolution afforded by the Herschel  Space Observatory at far-infrared and submillimeter wavelengths , we aim to characterize the physical properties of cold dust within nearby galaxies , as well as the associated uncertainties , namely the robustness of the parameters we derive using different modified blackbody models . For a pilot subsample of the KINGFISH key program , we perform two-temperature fits of the Spitzer  and Herschel  photometric data ( from 24 to 500 \mu m ) , with a warm and a cold component , both globally and in each resolution element . We compare the results obtained from different analysis strategies . At global scales , we observe a range of values of the modified blackbody fit parameter \beta _ { c } ( 0.8 to 2.5 ) and T _ { c } ( 19.1 to 25.1K ) . We compute maps of our modelling parameters with \beta _ { c } fixed or treated as a free parameter to test the robustness of the temperature and dust surface density maps we deduce . When the emissivity is fixed , we observe steeper temperature gradients as a function of radius than when it is allowed to vary . When the emissivity is fitted as a free parameter , barred galaxies tend to have uniform fitted emissivities . Gathering the parameters obtained each resolution element in a T _ { c } - \beta _ { c } diagram underlines an anti-correlation between the two parameters . It remains difficult to assess whether the dominant effect is the physics of dust grains , noise , or mixing along the line of sight and in the beam . We finally observe in both cases that the dust column density peaks in central regions of galaxies and bar ends ( coinciding with molecular gas density enhancements usually found in these locations ) . We also quantify how the total dust mass varies with our assumptions about the emissivity index as well as the influence of the wavelength coverage used in the fits . We show that modified blackbody fits using a shallow emissivity ( \beta < 2.0 ) lead to significantly lower dust masses compared to the \beta < 2.0 case , with dust masses lower by up to 50 \% if \beta _ { c } =1.5 for instance . The working resolution affects our total dust mass estimates : masses increase from global fits to spatially-resolved fits .