UDMS: A Canonical Schema and Assessment Framework for Measuring and Maintaining DJ Metadata Quality Across Ecosystems

DJ software libraries contain rich metadata (BPM, musical key, genre, ratings, labels) critical for harmonic mixing, playlist curation, and music information retrieval. However, metadata quality within real-world DJ libraries has not been systematically studied. We present the first cross-platform metadata quality analysis using UDMS, a Unified DJ Metadata Schema designed for cross-platform interoperability.

Analyzing 636 Rekordbox tracks, 382 Serato tracks, 112 Engine DJ tracks, and 190 VirtualDJ tracks, we measure field coverage across 11 canonical metadata fields. Audio-derived fields (duration, bitrate, sample rate) achieve 100% coverage and musical key achieves 93–98% across platforms, but genre annotation is incomplete for 31–69% of tracks and editorial metadata (ratings, labels) is sparse (0–53%).

Across 143 tracks present in both Rekordbox and Serato, title/artist/album are perfectly preserved (100%). BPM disagreement is explained in part by a systematic 2× half-tempo interpretation difference in Rekordbox affecting 27% of tracks; Engine DJ confirms the canonical tempo convention. Key achieves 71.3% exact / 100% effective agreement. A preliminary 24-track Rekordbox–VirtualDJ comparison suggests a distinct 1.19× algorithmic pattern. Genre is the most degraded field across platforms (79.1% agreement when both tagged), reflecting culturally-rooted taxonomy fragmentation rather than acoustic distinctions.

UDMS enables both diagnostic measurement and active maintenance: cross-platform BPM voting corrects systematic errors, multi-source key aggregation covers gaps no single platform closes alone, and per-track quality scoring provides continuous library health monitoring.

Open-source implementation and materials: https://github.com/interfluve-wav/dj-metadata-study