Nowadays, increasing demands in monitoring, control, and connectivity often require quantifying variables that are not easy to measure. Soft metrology allows the development of routines that objectively quantify magnitudes that are subjective, difficult, or expensive to measure. (As shown in this chapter, soft metrology measures quantities related to human perception or those derived from abstract representations, and this is a new conceptual paradigm for comprehending metrology in contrast to the well-known basic measurements directly associated with the International System of Units (SI).) However, despite the increase in the use of soft metrology systems, it is a developing area, and there are still many issues related to ensuring the validity of results, as there are no standardized uncertainty estimation procedures for these kinds of applications, and there still are open discussions about the uncertainty propagation through blocks of measurement processes made up by computer routines. This book chapter discusses the state of conceptual development and the proposed general structure for soft metrology systems, focusing on challenges involved in uncertainty estimation. Lastly, some discussions and final considerations are presented where epistemic and aleatory uncertainties are contrasted and associated with the representation quality and learning capability for inference, regression, and/or forecasting models in the framework of measurement processes and soft metrology.