Syndicated loan data provided by DealScan is an essential input in banking research. This data is rich enough to answer urging questions on bank lending, e.g., in the presence of financial shocks or climate change. However, many data options raise the question of how to choose the estimation sample. We employ a standard regression framework analyzing bank lending during the financial crisis of 2007/08 to study how conventional but varying usages of DealScan affect the estimates. The key finding is that the direction of coefficients remains relatively robust. However, statistical significance depends on the data and sampling choice and we provide guidelines for applied research.