In the previous chapters, the system identification loop was followed for two different flow control strategies. Using two data sets for each strategy, one for identification and another for validation, Reduced Order Models (ROMs) were computed and validated. This process used several tools, such as inbuilt MATLAB functions, which have their own underlying assumptions. The first aim of this chapter is to discuss such underlying assumptions, more specifically the initial conditions used in the linear simulations and the methods used to estimate the latter. These will be explored, along with the implications of using different methods. Furthermore, the models were used mainly for the first two applications of data-driven models discussed in Chap. 3 : to evaluate underlying dominant spatio-temporal patterns (diagnostics) and to use these to build models capable of predicting the system future behaviour (future state prediction). However, for more ambitious applications, such as control, other requirements may take place. For example, the ROM must be stable. For other applications, such as Model Predictive Control (MPC), a certain degree of causality may make it simpler to formulate the MPC problem mathematically. The methods discussed in this book do not ensure that the final ROM is stable, for example. There are, nevertheless, techniques capable of circumventing this challenge (up to a certain degree). This chapter also discusses these techniques, which were already introduced in Sect. 3.7 , and uses examples to discuss its implications. Because these topics are related to the nuances of modelling and simulation, these are referred to as advanced topics.