# Hierarchical Tagging for Kafka: Cost Attribution, Ownership, and Governance

Over 40% of organizations suffer from poor data quality, affecting their operations and decisions ([source: Experian](https://www.edq.com/)). A simple fix exists: tagging resources.

This article covers how tags save resources and improve data analytics accuracy.

## Tags Are Free-Form Metadata

A tag is a label assigned to a piece of information. Hashtags on X or Instagram work the same way. Click [#apachekafka](https://twitter.com/search?q=%23apachekafka) and you find everyone talking about it.

Trello and Linear use the same concept. Columns are tags that organize tickets.

![Screenshot of Trello and Linear boards using columns as tags to organize tickets](https://www.conduktor.io/assets/images/blog/trello-linear-tags-columns-example.png)

A tag on a resource can represent:

- Owner, team, project name
- Environment, update frequency (millis, daily, weekly)
- Internal code name, cost center

Tags can encode time as `key:value` pairs: `shutdown_at:xxx`, `restart_every_ms:xxx`. They trigger automation and GitOps workflows. Adding or removing a tag can kick off any process.

**Tags are free fields to store metadata on a resource.** A key-value store attached to anything.

In Conduktor, tags on topics clarify ownership and simplify management:

![Conduktor topic view showing tags labeling owner, team and environment to clarify ownership and management](https://www.conduktor.io/assets/images/blog/conduktor-topic-tags-ownership-management.png)

## Tags as Lossy Compression

Tagging is lossy compression for data categorization.

Data tagged as "PII" indicates personal identifiers like email, name, and phone number. In libraries, categorizing books under "A-B" doesn't pinpoint exact locations but enables binary-tree-style searching.

![Illustration of tagging as lossy compression: data labeled 'PII' narrows the search like a library shelf grouping books by alphabet range](https://www.conduktor.io/assets/images/blog/kafka-tagging-pii-search-illustration.png)

Tags offer multiple perspectives on data, like a prism. They simplify resource selection for analysis, organize information, and support visual representations like charts and tables.

## Tags Enable Data Mesh

Conduktor believes in Data Mesh principles. Our mission is helping platform teams adopt this model to maximize their data's value.

In Data Mesh, tags categorize everything into Data Products. They define:

- Ownership, description
- Semantic models, data schemas
- SLAs, lifecycle, intended use
- Data quality, freshness, lineage, versioning, origin

![Diagram showing tags on a data product capturing ownership, semantic models, SLAs, data quality, freshness, lineage and versioning](https://www.conduktor.io/assets/images/blog/kafka-tagging-data-quality-freshness-lineage.png)

Tags form a core component of data governance. Most are curated manually, though AI is moving toward automating suggestions.

Missing tags signal poor data quality and weak governance. Tags don't just organize data. They add **context** and **meaning**, improving accessibility, compliance, and utility across the ecosystem.

In Data Mesh architecture, where decentralized ownership and domain-oriented design matter, tags act as connective tissue. They ensure coherence in sprawling data landscapes.

## Tags Enable ABAC Security

Attribute-Based Access Control (ABAC) is an alternative to RBAC. Permissions depend on attributes (tags) of resources and users.

Tags apply to roles and users, not just resources. This enables ABAC policies based on "who" the user is and "what" they access. Authorization happens only when user tags match resource tags.

**Example**: Sensitive data tagged `sensitive` requires users or roles tagged `can_sensitive` to interact with it.

![Diagram of ABAC policy: data tagged 'sensitive' only accessible to roles or users carrying the 'can_sensitive' tag](https://www.conduktor.io/assets/images/blog/kafka-sensitive-tag-role-policy-example.png)

## Hierarchical Tags Beat Flat Lists

Tags can be hierarchical, like folders. Nested structures clarify relationships between concepts.

In large ecosystems with many tagged items, hierarchical structures outperform flat tagging. They simplify navigation and reduce conflicts.

In diverse organizations, the same tag may mean different things across teams. Combining flat tags to understand the big picture gets messy.

Hierarchical tagging introduces order and intuitive data relationships:

- `roles/owner`
- `sales/onlinesalesteam/summary`
- `geo:europe/france/paris`

## Beyond Tags: Ontologies

Not the metaphysical kind. Data ontologies are structured frameworks for organizing information that define categories, properties, and relationships.

Ontologies carry more meaning than tags. AI systems, databases, and the semantic web use them when rich, formal structures are necessary.

Ontologies establish a comprehensive, nuanced, and interconnected data model beyond basic categorization.

## Tag Everything, Everywhere

Tags are essential for any serious data ecosystem. Their purpose varies by scale: metadata, automation, analytics, or compliance.

Whether you're a team of 5 with hundreds of resources or 100 people with five resources, tagging everything pays off.

Conduktor helps tag resources like topics for classification, search, and ownership. Tags provide quick insights into your ecosystem and enable faster decisions through automation.