In this study, we modelled the density of electricity smart meter data using different implementations of the KD and CKD estimators. The methods were aimed at accommodating the seasonality in consumption, along with the underlying variability. We used a decay parameter in the modelling framework to place more emphasis on the more recent observations. The evaluation of post-sample density and quantile forecasts showed that the methods considered in this study convincingly outperformed the unconditional KD estimator. Encouragingly, the employed methods were able to accommodate the weekly seasonality that can be present in consumption. We also implemented the HWT exponential smoothing method, and found that point and density forecasts from this method were particularly competitive for the SME series. This can be attributed to the relatively strong seasonal patterns typically present in the SME series. Furthermore, we derived density forecasts for electricity cost using the density forecasts of electricity consumption, for six different tariffs. Using three different tariff selection criteria, our empirical study showed that switching between tariffs results in considerable cost savings overall, compared to the case when consumers are allocated a single tariff for all periods