Process monitoring in technology and industry in general, in pharmaceutical batch and continuous manufacturing in particular, is incomplete without full understanding of all sources of variation. Pharmaceutical mixture heterogeneity interacts with the particular sampling process involved (by physical extraction or by Process Analytical Technology (PAT) signal acquisition) potentially creating four Incorrect Sampling Errors (ISE), two Correct Sampling Errors (CSE) in addition to the Total Analytical Error (TAE). In the highly regulated pharmaceutical production context it is essential to eliminate, or reduce maximally, all unnecessary contributions to the Total Sampling Error (TSE) to the Measurement Uncertainty (MU_total) in order to be able to meet stringent regulatory blend and dose uniformity requirements. Current problems mainly stem from inadequate understanding of the challenges regarding sampling of powder blends. In this endeavor the Theory of Sampling (TOS) forms the only reliable scientific framework from which to seek resolution. We here present the variographic approach with an aim to conduct TSE error variance identification and to show how to develop fit-for-purpose acceptance levels in critical powder blending process monitoring. The key issue regards the nugget effect, which contains all non-optimised [ISE, CSE] plus TAE contributions to MU_total. A large nugget effect w.r.t. the sill is a warning that the measurement system is far from fit-for-purpose, and must be improved. Regulatory guidances have hitherto called for physical sampling from within blenders, leading to significant ISE associated with the insertion of sample thieves (sampling spears). Instead of self-crippling spear sampling we here call for a paradigm shift, very much from the TOS regimen, in the form of alternative on-line variographic characterisation of 1-D blender outflow streams. Practical illustrations and case histories are described in parallel contributions to WCSB7.