To produce accurate and artefact-free MTDSC data it is necessary to choose suitable experimental parameters (i.e. modulation period, temperature amplitude, sample mass and pan type). A poor choice of parameters can result in distortion of the modulation and inaccurate heat capacity data and can also affect the result of the deconvolution procedure. This paper aims to review these effects and to offer some guidance on how to achieve optimum results. In particular, it is suggested that the modulation parameters should be chosen to ensure that the sample can follow the programme without thermal gradients developing and also that at least six modulations should take place during any thermal transition. Two methods of calibration are discussed; a one-point method is sufficient in most cases but a temperature-dependent method should be used when more accurate data are required or for sub-ambient experiments.