<p>One of the primary inconveniences of the conventional Brazilian test for determining the indirect tensile strength of brittle materials is the influence of the contact chord length on the stress state generated at both the center of the specimen and the actual failure initiation point. To overcome this problem, the flattened Brazilian test, in which the total contact length is supposed to be fixed beforehand, has been proposed for determining the indirect tensile strength of rocks and concrete. In this study, the actual distribution of contact stress between the loading platens and the specimen is characterized through numerical simulations, revealing a non-uniform stress distribution. Furthermore, the study quantifies the influence of non-parallelism between the loaded faces, showing that specimens with parallelism deviations below 0.05° tend to fail at the center, despite the presence of secondary cracks. In contrast, deviations within the range of 0.05– 0.10° generate secondary cracks near the contact, which for lacks of parallelism above 0.10° may result in failure initiation near the contact region. Based on these findings, two novel contact stress distributions are proposed: a standard parabolic distribution and a shifted parabolic distribution. The resulting formulations more accurately replicate the observed stress conditions and account for the effects of geometric imperfections. Both numerical and analytical results are validated through real testing using digital image correlation (DIC) techniques. Consistency among analytical predictions, numerical simulations, and DIC results proves that flatness geometric deviations trigger failure at locations closer to the contact rather than near the specimen”s center, even within the conventionally accepted contact angle range of 20° to 30°. Actually, this range proves to be dependent on the shape of the contact distribution generated along the loaded boundaries.</p>