The combination of indirect test and built-In Self-Test (BIST) is a promising solution to mitigate the increasing cost of testing complex mixed-signal integrated systems. Indirect test replaces complex specification measurements by simpler signatures, and then uses moden data analysis algorithms to map these signatures onto the specification space. Signatures can be efficiently monitored by simple on-chip test instruments that can be integrated together with the system under test. Indirect test is then an interesting path to enable cost-effective BIST for mixed-signal systems. This PICS project has the goal of developing reliable and accurate built-in indirect test methods for complex mixed-signal systems. The project is structured into two interconnected research lines: a) Combining causal inference techniques with feature selection anda feature extraction algotirhms for indirect test, and b) Developing a feature-driven strategy for the definition of on-chip test instruments.