In spite of intense efforts in the last 20 years, the current state of affairs regarding evaluation of adequacy of pharmaceutical mixing is at an impressive standstill, characterized by two draft guidances, one withdrawn, and the other never approved. We here analyze the regulatory, scientific and technological situation and suggest a radical, but logical approach calling for a paradigm shift regarding sampling of pharmaceutical blends. In synergy with QbD/PAT efforts, blend uniformity testing should only be performed with properly designed sampling that can guarantee representativity-in contrast to the current deficient thief sampling. This is necessary for suitable in-process specifications and dosage units meeting desired specifications. The present exposé shows how process sampling based on the Theory of Sampling (TOS) constitutes a new asset for regulatory compliance, providing procedures that suppress hitherto adverse sampling errors. We identify that the optimal sampling location is after emptying the blender, guaranteeing complete characterisation of the residual heterogeneity. TOS includes variographic analysis that decomposes the effective total sampling and analysis error (TSE+TAE) from the variability of the manufacturing process itself. This approach provides reliable in-process characterization allowing independent approval or rejection by the Quality Control unit. The science-based sampling principles presented here will facilitate full control of blending processes, including whether post-blending segregation influences the material stream that reaches the tabletting feed-frame.