Abstract Dimensional variation in assembly processes is one of the most important issues that affect quality. Although robust design and statistical process quality control help to reduce this problem, they cannot be used for instant variation reduction during assembly operations, especially during process ramp-up. This paper introduces a complete methodology for dimensional-related error compensation in compliant sheet metal assembly processes. The proposed methodology is divided into two steps: (1) an off-line error control-learning module using virtual assembly models to determine necessary adjustments; and (2) an in-line control implementation using a feed-forward control strategy based on the learned adjustments. The off-line learning step focuses on determining control actions or corrections to compensate for the negative effects incoming part errors have on Key Product Characteristics. Specifically, it utilizes a newly developed iterative sampling method based on Kriging fitting to efficiently determine optimal control actions. The in-line feed-forward control identifies appropriate part-by-part adjustments using these learned control actions and incoming assembly component measurements. In this paper, two case studies are presented. First, a mathematical case study presents an empirical proof for the feasibility of the Iterative Sampling and Fitting Algorithm. Second, a simulation-based case study illustrates the effectiveness of the proposed methodology to improve dimensional quality in assembly operations for compliant sheet metal parts.