ADAPT-Select: Reference Implementation and Reproducibility Artifact

Abstract

Reference implementation and reproducibility artifact for the manuscript "ADAPT-Select: A Gated Multi-Criteria Algorithm for Recommending Modern Data Architectures." ADAPT-Select is a gated semi-compensatory decision algorithm for choosing among five data-platform paradigms (warehouse, lake, lakehouse, mesh, fabric). A conjunctive feasibility gate removes paradigms an organization cannot operate, compensatory weighted-additive scoring ranks the survivors, and a coverage optimizer returns a minimal hybrid composition. Two further stages apply the same gating semantics to cross-layer tool compatibility and to budget feasibility under a FinOps remediation ladder. Contents:• reference-implementation/ — dependency-free installable Python package (library and CLI) implementing all five stages, with a test suite asserting Propositions 1, 6 and 7 and reproducing the paper's worked example• adapt_select.py — paradigm-level simulation: 20,000-profile generator, maturity gate, and the SAW, TOPSIS, PROMETHEE II and lexicographic baselines• stage4_sim.py, stage5_sim.py — stack-compatibility and FinOps cost-remediation simulations• results_alpha1.json, results_stage4.json, results_stage5.json — outputs regenerating every numerical result reported in the paper from a fixed seed• fieldkit/ — Best-Worst Method solver and elicitation materials for the validation protocol• examples/ — four archetype organizational profiles All reported rates are internal-consistency measurements over expert-synthesized models, not field measurements. The simulation scripts require NumPy, SciPy and Matplotlib; the reference implementation package has no runtime dependencies.