Both system codes and experiments are used for simulating nuclear reactor thermal-hydraulic transients in safety analysis. System codes model the whole reactor and must demonstrate their quality by extensive validation against experimental data, which cover the important phenomena. Integral effect test facilities intend to simulate reactor thermal-hydraulic behaviour in reduced scale conditions. Advanced scaling methods exist to define how to respect the dominant phenomena in a scaled experiment and to evaluate distortions. They perform a similarity analysis applied to equations governing the evolution of important parameters at system, component and local levels. They require an evaluation of the order of magnitude of every term of these equations. Today, system codes have reached a good maturity and reliability, and one defines in this work the conditions for a proper use of codes for evaluating terms of the equations used in scaling analyses. It appears that scaling tools at system level can help in the analysis of code predictions, in the establishment of a structured PIRT and for identifying the sources of code predictions differences with experimental data. This paper presents the lessons learnt from a combined use of scaling methods, scaling tools, system codes and integral effect tests data. Some conclusions and recommendations are drawn on the merits of the various scaling methods, on the selection of the equations to be used in the scaling analyses, on the various methods for estimating terms of the selected equations, and on the acceptability limits for the distortions. The applications refer to some PWR LOCA analyses.