In rare disease drug development where no registry or natural history dataset exists, real-world evidence quality depends on treating evidence engineering as a design-stage decision, defining intended regulatory use and standardizing endpoints, harmonization, and data provenance upfront rather than reconciling gaps after collection.