A process operating performance optimality assessment (POPOA) consists of an optimal degree online assessment and non-optimal cause identification, which contribute to maintaining a high comprehensive economic index (CEI) of the production. However, two main problems limit the application of the traditional POPOA methods, i.e., the plant-wide process characteristics and the coexistence of both the quantitative and qualitative variables. To overcome the two problems for POPOA, a novel two-level multi-block assessment method based on the fuzzy probabilistic rough set (FPRS) is proposed in this research. The operating performance grade of both the global and sub-block level are properly defined, where the sub-block assessment indices, which are difficult to obtain, are not required. Different from traditional multi-block methods due to the novel offline modelling method, an explicit global model is unnecessary. The global performance grade is directly determined by the sub-block performance grades. When the process is operating at a non-optimal performance grade, the responsible sub-block can be rapidly identified through online assessment. The proposed non-optimal cause identification technique is carried out in the non-optimal sub-blocks, based on a newly-defined matching degree function. The identified non-optimal causes also contribute to the actual production adjustment to obtain the optimal performance. Finally, the proposed POPOA method is successfully applied to a gold hydrometallurgy process, which is a typical plant-wide process with hybrid types of variables.