Software inspections have shown to contribute to improved software quality. However, as they are time consuming, sometimes the resources available are not sufficient to inspect all documents. Instead of selecting documents ad hoc, Sample-Driven Inspections (SDI) is proposed as a systematic approach to select which subset of documents to inspect. The selection is based on a pre-inspection, where samples of documents are inspected, and based on the sample, it is estimated which documents need the quality improvement the most. Two important questions are how to select the pre-inspection sample, and how large segments the documents should be divided into. In this paper, we apply the SDI approach to industrial data from inspections of requirements specifications. Different sampling strategies are applied and the documents are split into segments of different size. It is concluded that the SDI approach is more effective in most cases compared to selecting subsets of documents randomly, despite the fact that the faults are rather evenly distributed in the documents.