In this paper, we develop a data-driven model of a model-sized aircraft using system identification (SysID) techniques. The emphasis is placed on multiple short data records that are used in obtaining an initial model of the system. The records are 'short' with respect to the length of a 'typical' identification experiment and are necessary because of the unstable nature of the open-loop system. Owing to this, our formulation can also be applied to other systems that are highly non-linear or have external influences that act over relatively longer periods of time. The length of each record is at least an order smaller than what a full-length PseudoRandom-Binary-Sequence (PRBS) sequence would require. For the model aircraft, the lengths are a second or less and are dictated by the length of time a 'novice' operator could keep the aircraft in unconstrained flight. A set of decentralized PID controllers are then tuned using robust control techniques augmented with computation of control-relevant uncertainty estimates from data. The uncertainty, together with closed loop sensitivities serve as an essential metric to evaluate the utility of the obtained model. The designed controller showed acceptable stabilization properties for the aircraft in hover mode. A second iteration of SysID using longer data records from the initial stabilized system resulted in a controller with improved bandwidth and reference tracking. Results from the experiment are presented to illustrate the sequence of operations.