The idea of accuracy extrapolation was launched in this paper. The concerned area is nuclear thermal-hydraulic and the issue is the capability to predict large Nuclear Power Plant (NPP) transient performance when only small scale experiments are available. It is clear that experimental data measured in differently scaled facility (whatever is their size and operating conditions) cannot be extrapolated to NPP conditions due to a number of well recognized scaling distortions (like heat losses, thermal energy release from passive structures, etc.). Furthermore one may not scale-up results from code calculations because of numerical issues (results affected by node size and relative position of nodes). However one may scale up the error or the accuracy of results, i.e. the combination, for instance the ratio, of measured and calculated values. The accuracy may become available from facilities (and related calculations) of different size and scale and may be extrapolate to the NPP (prototype) condition.