1. The document presents a technique called SATT (Software Architecture-based Threshold Tailoring) that tailors code metric thresholds to different software architectures.
2. SATT works by ordering classes by their metric values, weighting them based on their size, aggregating the weights to extract thresholds at certain points, and defining thresholds specific to architectural roles whose metrics differ significantly from other classes.
3. The study applies SATT to projects using common architectures like MVC and finds that architectural roles do have significantly different metric value distributions compared to other classes, which can be explained by their specialized responsibilities. SATT improves over benchmark-based techniques by setting role-specific thresholds.