StanIa posterior PDFs of w for different cosmological sample sizes (line styles) from the DDF and WFD observing strategies (panels) under the ‘perfect’ (black), ‘random’ (magenta), and ‘fiducial’ (cyan) mock classification schemes, with the true value of w indicated by a brown vertical line. We note that the ‘random’ classifier’s constraints in the DDF underestimate w by so much that they cannot be shown on these axes. As the constraints are not very sensitive to the sample size, we can safely use a cosmological sample of 3000 light curves in our tests.