Continuous manufacturing is now considered as a well-established technique by pharmaceutical companies. However, the limited number of filed applications reflects the complexity to translate a science that has been described in many publications to an actual drug product. Process stability evaluation and resulting sampling and diversion strategy are key aspects of the design of continuous processes which require the development of new approaches. This study describes a new methodology to evaluate process stability for a continuous line based on twin-screw granulation. In such lines, both continuous and discrete unit operations are present. The diversion and quality decision of intermediate product is therefore made at the level of individualized portions of the batch called product keys (PK). The described methodology therefore evaluates the process stability at PK level. A batch statistical process model was calibrated with three manufacturing campaigns and verified on five independent campaigns. The developed model allowed identifying outlying PKs within a manufacturing campaign. This approach gives new perspectives for rationalizing the sampling strategy, designing the diversion strategy and continued process verification. Further extension of the model could be considered to enable its use for quality decision.