If properly calibrated, the shapes of galaxy clusters can be used to investigate many physical processes: from feedback and quenching of star formation, to the nature of dark matter. Theorists frequently measure shapes using moments of inertia of simulated particles’. We instead create mock (optical, X-ray, strong- and weak-lensing) observations of the twenty-two most massive (∼1014.7 M⊙) relaxed clusters in the BAHAMAS simulations. We find that observable measures of shape are rounder. Even when moments of inertia are projected into 2D and evaluated at matched radius, they overestimate ellipticity by 56% (compared to observable strong lensing) and 430% (compared to observable weak lensing). Therefore, we propose matchable quantities and test them using observations of eight relaxed clusters from the Hubble Space Telescope and Chandra X-Ray Observatory. We also release our HST data reduction and lensing analysis software to the community. In real clusters, the ellipticity and orientation angle at all radii are strongly correlated. In simulated clusters, the ellipticity of inner (<rvir/20) regions becomes decoupled: for example with greater misalignment of the central cluster galaxy. This may indicate overly efficient implementation of feedback from active galactic nuclei. Future exploitation of cluster shapes as a function of radii will require better understanding of core baryonic processes. Exploitation of shapes on any scale will require calibration on simulations extended all the way to mock observations.