Impact of unmodelled systematic biases on the posterior of a likelihood analysis with heatmaps. In all cases, we do not marginalise over any systematic effects. The target data vector is then infused with one systematic bias, and we run a likelihood analysis for this infused data vector. For comparison, we show the constraints on a data vector that is not infused by systematics (grey). Note that the values of the dz shifts are given in units of the standard deviation of the dz prior (compare Table 1).