Robust optimization: Lessons learned from aircraft routing

•We model uncertainty in composite variable models under extreme-value, chance-constrained and domain-specific paradigms.•We show the lack of pareto-optimality in existing models and develop advanced versions for modeling ease and tractability.•Robust models are found to be superior in terms of on-time performance, propagated delays and passenger disruptions.•We show that the efficacy of a robust approach is determined by the interaction between model, data and evaluation metrics.•The findings and extensions from this work are applicable to a broad class of network-based resource allocation problems.