Most cosmological data analysis today relies on the Friedmann-Lemaitre-Robertson-Walker (FLRW) metric, providing the basis of the current standard cosmological model. Within this framework, interesting tensions between our increasingly precise data and theoretical predictions are coming to light. It is therefore reasonable to explore the potential for cosmological analysis outside of the exact FLRW cosmological framework. In this work we adopt the general luminosity-distance series expansion in redshift with no assumptions of homogeneity or isotropy. This framework will allow for a full model-independent analysis of near-future low-redshift cosmological surveys. We calculate the effective observational 'Hubble', 'deceleration', 'curvature' and 'jerk' parameters of the luminosity-distance series expansion in numerical relativity simulations of realistic structure formation, for observers located in different environments and with different levels of sky-coverage. With a 'fairly-sampled' sky, we find 0.6% and 4% cosmic variance in the 'Hubble' and 'deceleration' parameters for scales of 200 Mpc/h (corresponding to density contrasts of ~0.05 in the simulated model universe), respectively. On top of this, we find that typical observers measure maximal sky-variance of 2% and 120% in the same parameters, as compared to their analogies in the large scale FLRW model. Our work suggests the inclusion of low-redshift anisotropy in cosmological analysis could be important for drawing correct conclusions about our Universe.

20 pages, 12 figures, v2: matches published version in Phys. Rev. D, v3: updated results to match erratum