Semiconductor manufacturing is highly complex and expensive, hence the early detection of problems is necessary to minimize the number of scraps and improve the overall yield. This paper presents an industrial application of dynamic sampling based on an aggregated risk indicator at process tool level. The objective is to identify the lots that should be measured to minimize the overall risk level of the fabrication plant (fab). Results show significant improvements compared with the previous strategy: Sampling decisions are better adapted to the current production state and to the workload in the inspection area. Several parameters and algorithms are proposed and compared using industrial data.