MembershipType, LinkVerification, and Confidence in Cluster-Based Resource Identification

In cluster‑based resource identification systems, multiple data points are often used to justify the inclusion of a particular resource within a cluster. Three key elements — MembershipType, LinkVerification, and Confidence — work together to provide transparency, provenance, and assurance regarding the accuracy of each ClusterMember relationship. These elements form a dialectic set of indicators that help both senders and recipients assess the trustworthiness of a cluster and make informed operational decisions.

These attributes serve to:

  • Explain why a resource has been included in a cluster;

  • Provide insight into the method used to validate the association;

  • Indicate the sender’s level of certainty;

  • Support disambiguation between similar or identical resources; and

  • Enable recipients to assess the overall trust level of a cluster.

Although these elements may complement other metadata or governance processes, they can also stand alone as a framework for assessing evidential strength in cluster formation and exchange.

MembershipType

Indicates the basis on which the cluster was formed. Typical sources include:

  • Audio reference files;

  • Reference metadata; and

  • A combination of both.

This element helps recipients understand the nature and reliability of the inputs used to justify the association.

LinkVerification

Describes the level of validation performed to support the inclusion of a resource as a ClusterMember. It provides a clear signal of the checks and balances applied.

Common values include:

  • VerifiedByHuman – a person has reviewed and confirmed the linkage;

  • CrossChecked – multiple sources or processes corroborate the association;

  • NotVerified – no verification actions were taken; and

  • UserDefined – an organisation-specific verification method was applied.

Confidence

Expresses the sender’s degree of certainty as a percentage that a resource has been correctly associated with the cluster.

Confidence helps recipients assess the probability that the inclusion is correct and can be used to rank or filter associations especially when see in the context and relationship of the MessageSender.

Additional Elements Supporting ISRC Disambiguation and Trust

To enable the recipient to assess whether the ClusterCreator has correctly determined that multiple ISRCs identify the same underlying sound recording according to the applicable ISRC disambiguation criteria, the message provides the full set of ISRC reference metadata elements. Of these, only ResourceId, ResourceType, Title, DisplayArtistName, and Duration are mandatory. However, the inclusion of additional reference metadata significantly improves the disambiguation and matching process for the message recipient. In general, the more metadata elements that are supplied, the stronger the evidentiary basis for evaluating the asserted relationship between ISRCs.

In addition to the ISRC reference metadata defined in the ISRC Handbook, the ClusterMember composite includes supplementary metadata elements that further support accurate disambiguation of sound recordings. Although these additional metadata elements are optional, their presence can materially strengthen the recipient’s confidence in the clustering decision by providing deeper contextual and technical information. In particular, the following elements offer enhanced support:

  • RecordingMode describes how the audio was captured or processed;

  • Fingerprint composite outlines the algorithmic or audio‑based identifiers used for matching; and

  • MessageSender and ClusterCreator which indicate the entity responsible for the clustering decision and the segment of the music value chain in which they operate (e.g., rights holder, MRT provider, distributor, or other stakeholder). This provides essential context regarding the sender’s typical access to authoritative data.

These provide further context and help recipients understand the provenance of the clustering decision.

In summary, the overall trust level of a cluster can be evaluated by combining:

  • The MessageSender of the message;

  • The ClusterCreator of the cluster;

  • The Confidence score;

  • The LinkVerification value; and

  • The MembershipType.

Together, these form a holistic trust‑assessment model that supports informed decision‑making when exchanging resource clusters across organisations.

Search