During multiparameter chemical optimization, molecular capabilities increase as hits evolve into leads and development candidates. Like retrosynthetic analysis, where target molecules are transformed into structurally simplified starting materials, we introduce retro-optimization analysis, transforming sophisticated development candidates into less capable leads and hits. To retrospectively understand the logic of optimization in discovery campaigns, we enumerated a matched molecular pair network and compared the actual route of optimization taken to alternative theoretical routes of optimization. We noted differences in the network and properties of the lead molecule compared to those of alternatives. We identified substructures of the evolving ligand, named "optimizons," and tracked their emphasis and discovery. While we initially defined and illustrated these methods for a single project, our expansion to three additional discovery projects and three external data sets proved consistent. We retrospectively define the logic of optimization at the project, molecular, and submolecular levels to prospectively guide current and future optimization campaigns.