Deduplicating and Merging Sound Recording Registrations Without a Reliable ISRC
In short: Noctil's MatchIQ engine finds every registration of the same sound recording, even when the ISRC is missing, duplicated or wrong. It clusters them against each other and against the organization's existing repertoire. ProcessEngine then merges each cluster into one complete recording, with its full performer line-up and rights ownership, under the organization's own rules, sending only uncertain cases to human review. In a 100,000-pair benchmark, recording-to-recording matching produced fewer than 0.01% false positives or false negatives. 98% of duplicate registrations were clustered and merged automatically, and manual review fell by up to 35%.
Turn many registrations of the same recording into one complete, trusted record. For neighbouring rights societies, performer and producer CMOs, and record label collecting bodies that receive the same sound recording from several labels, distributors, performers and partner societies.
The Situation
Neighbouring rights societies collect for the performers and producers of sound recordings. They receive recording data from many directions: record labels and their distributors, featured and non-featured performers and their agents, rights administrators, and partner societies abroad.
Each source registers the recordings it has an interest in, at different times and in different formats. One track can reach the society as a label registration, several performer claims and a partner society’s repertoire file. Each one carries its own partial view of the performer line-up, the rights owner and the release details.
The Problem
Combining those registrations into one accurate recording is harder than it looks:
- ISRCs can’t be relied on alone. Many registrations, especially performer claims and older catalog, arrive without an ISRC. Others carry the wrong one, or the same recording appears under several ISRCs after reissues and re-deliveries.
- Recordings versus releases. The same recording appears on albums, singles, compilations and reissues, each with its own UPC or EAN and track details. Registrations tied to different releases have to be resolved to one recording.
- Versions that must stay separate. Remasters, radio edits, live recordings, remixes and re-recordings can look nearly identical but are distinct recordings with different rights. Merging them by mistake misdirects royalties.
- Conflicting performer line-ups. Labels list featured artists, while performers claim session and non-featured roles. Names, instruments and roles differ between sources, and the full line-up only emerges once they are combined.
- Ownership by territory. The rights owner of a recording can differ by country and change over time through licensing and catalog sales. That has to be captured accurately, not overwritten.
A wrongly merged or incomplete recording leaves performers out or credits the wrong rights owner. That delays royalties, or pays them to the wrong people.
The Solution
The Noctil Platform combines MatchIQ for matching with ProcessEngine for workflow, to build one source of truth for every sound recording.
- Ingest from every source. Registrations from labels, distributors, performers, administrators and partner societies are loaded in their native formats, including DDEX.
- Match and cluster with MatchIQ. MatchIQ deduplicates incoming registrations against each other and against the existing repertoire. It uses title and version, main and featured artists, performers and roles, duration, label, release details, and ISRC and UPC/EAN where present. Every registration of the same recording is linked into one cluster, even when the ISRC is missing or wrong.
- Version-aware resolution. Remasters, edits, live recordings and remixes are recognized as distinct recordings and kept separate. Reissues and compilation appearances of the same recording are linked to it.
- Rule-based merging. Each cluster is merged into a single recording using the organization’s own rules for source authority, performer line-up, rights ownership by territory, data quality and completeness.
- Workflow orchestration with ProcessEngine. ProcessEngine runs the end-to-end merge workflow. It sends clusters to human review only when confidence is low or performer and ownership claims conflict, and records every decision for audit.
- Continuous enrichment. New claims, releases and identifiers are matched to the existing recording and enrich it rather than creating duplicates. Alternate ISRCs are linked to the recording instead of creating new records.
The Result
- Fewer than 0.01% false positives or false negatives in recording-to-recording deduplication, based on a 100,000-pair matching benchmark
- 98% automatic clustering and merging of duplicate registrations in benchmark testing
- Up to 35% less manual review, through fewer candidate matches and an automated enrichment workflow
- Version-level accuracy, so every remaster, edit and live recording stays a distinct, complete record
- Complete performer line-ups and rights ownership on every recording, so royalties flow without delay
Frequently Asked Questions
How do neighbouring rights societies deduplicate sound recording registrations?
Noctil's MatchIQ compares each incoming registration with the others and with existing recordings, using title, version, artists, performers, duration, label and release details. It clusters every registration of the same recording, and ProcessEngine merges each cluster into one record.
Can recordings be matched without an ISRC?
Yes. MatchIQ matches on the combination of recording attributes, so registrations without an ISRC, or with a wrong one, are still clustered correctly. Alternate ISRCs are linked to the same recording.
Are remasters, edits and live versions merged together?
No. MatchIQ recognizes them as distinct recordings and keeps them separate. Reissues and compilation appearances of the same recording are linked to it.
How are conflicting performer claims handled?
Performer line-ups from labels and performers are combined under the organization's rules. Conflicting claims go to human review through ProcessEngine.
How accurate is recording-to-recording deduplication?
In a 100,000-pair benchmark, false positives and false negatives were below 0.01%, and 98% of duplicate registrations were clustered and merged automatically. Manual review fell by up to 35%.
Who is this for?
Neighbouring rights societies, performer and producer CMOs, record label collecting bodies, and rights administrators that maintain a sound recording repertoire built from many registration sources.