As more content creators embrace HDR workflows, the ability to consistently deliver high-quality visuals depends not only on the conversion process itself but also on understanding how to evaluate and optimize it effectively. In particular, content providers and streaming platforms are increasingly converting SDR (Standard Dynamic Range) content to HDR (High Dynamic Range) upstream—during encoding or streaming—using Look-Up Tables (LUTs), while TV manufacturers often handle this transformation on the device side. Our team developed a novel, feature-based quality metric purpose-built for SDR-to-HDR conversion workflows. This metric leverages machine learning models trained on expert-curated visual features, selected in close collaboration with experienced colorists. The result is a robust, content-aware system capable of ranking multiple LUTs according to their perceived visual quality.