Skill Hub — DAMA / Data
Metadata Management Working Skill
Catalog business, technical, and operational metadata. Identify gaps that block reporting, integration, or AI readiness. Produce a metadata inventory with ownership.
What this skill is for
This skill helps you document what your data means, where it lives, how it is structured, who maintains it, and how it is used. It produces a metadata inventory that resolves "what does this field mean?" disputes, unblocks integration projects, and prepares data for AI consumption. The output is a working catalog, not a theoretical model.
When to use this skill
- A report column has no documented definition and three departments give three different explanations.
- An integration project stalls because the source and target teams cannot agree on field semantics.
- A data scientist or AI agent cannot use a dataset because field meanings, formats, and update frequencies are unknown.
- A system migration requires mapping old fields to new fields, and no one knows what the old fields mean.
- An audit requires data lineage documentation and the existing documentation is outdated or missing.
- A new system is being integrated and the API schema lacks descriptions, examples, or ownership.
Real work situations
Example 1: The mystery report column
A BI report shows "Net Revenue." Finance says it is after discounts. Sales says it is before discounts. Operations says it excludes returns. No documentation exists. The skill produces a metadata entry for the column with a business definition, the calculation formula, the source tables, and the owner who can approve the definition.
Example 2: API integration confusion
A team builds an API to expose customer data. The field customer_status accepts values 1, 2, 3. The consuming team does not know what these values mean. The API documentation is a swagger file with no descriptions. The skill produces a metadata catalog for the API with business definitions, allowed values, examples, and update frequency.
Example 3: Migration mapping deadlock
A legacy system is being replaced. The migration team needs to map 400 fields. Half have no description in the legacy system. The skill produces a metadata discovery plan: interview legacy system experts, sample data for inferred meanings, document assumptions, and validate with business owners.
Example 4: AI readiness gap
An AI project needs structured data to train a model. The available datasets have no documentation on field meanings, data quality, update frequency, or ownership. The skill produces a metadata inventory that rates each dataset on AI readiness and flags gaps that must be closed before the dataset is usable.
Inputs required
- List of data elements, tables, APIs, or report columns in scope.
- System documentation, data dictionaries, or ER diagrams (even if outdated).
- Subject matter experts who know the business meaning of the data.
- Technical contacts who know the physical implementation (tables, fields, data types).
- Sample data for each element (to infer formats, ranges, and anomalies).
- List of consumers: reports, integrations, applications, or AI models that use the data.
- Existing metadata tools or catalogs (optional).
Questions to ask
- What does this field mean in business terms, not technical terms?
- Who uses this field, and what decision do they make with it?
- What are the allowed values, and what does each value mean?
- Where is this field physically stored: table, column, system, database?
- How often is this field updated, and by what process or user?
- What other fields or tables depend on this field?
- Who is the authoritative source for the definition of this field?
- What is the data quality of this field: completeness, accuracy, timeliness?
Working method
- Scope the metadata domain. Choose a bounded scope: one system, one data domain, one report, or one API. Do not attempt to catalog the entire enterprise at once.
- Identify metadata types needed. For each element, decide which types you need: business metadata (meaning, owner, rules), technical metadata (table, column, type, length), operational metadata (update frequency, source job, last change date), or usage metadata (consumers, reports, integrations).
- Collect existing documentation. Gather data dictionaries, ER diagrams, API specs, and wiki pages. Mark each source with a freshness date and trust level.
- Interview subject matter experts. Use the questions above. Record answers as facts or assumptions. If two experts disagree, document both views and flag for resolution.
- Sample and inspect data. Query the actual database or API to confirm physical names, data types, value distributions, and anomalies. Do not trust documentation alone.
- Document each element. For each data element, produce a metadata entry with: name, definition, type, source system, physical location, allowed values, owner, consumers, update frequency, and quality notes.
- Identify metadata gaps. Flag elements with missing definitions, missing owners, unknown consumers, undocumented values, or stale documentation. Classify each gap by business risk.
- Validate with owners and consumers. Ask the named owner to confirm the definition. Ask a consumer to confirm that the documented meaning matches their usage. Adjust if they differ.
- Produce the metadata inventory. Compile all entries into a single catalog. Include a gap summary, a risk assessment, and a closure plan.
- Set maintenance rules. Define who updates the catalog when systems change, how often the catalog is reviewed, and what triggers an update.
Decision rules
- If a field has no business definition, do not use it in a report or integration until the definition is documented and approved.
- If two systems use the same field name with different meanings, rename one or document the semantic difference explicitly.
- If a field's allowed values are undocumented, sample the data and present the inferred values to the owner for confirmation.
- If a field has no owner, flag it as a governance gap and assign an owner before the metadata entry is marked complete.
- If documentation contradicts the actual data, trust the data and flag the documentation as stale.
- If a metadata entry is only technical (no business meaning), it is incomplete; add business metadata before marking it done.
- If a dataset has no documented consumers, question whether it should be maintained; unused data has a cost.
Deliverables
- Metadata Inventory — catalog of data elements with business, technical, and operational metadata.
- Metadata Gap Report — list of missing or stale metadata with business risk and closure actions.
- AI Readiness Assessment — rating of datasets on metadata completeness, quality, and ownership.
Templates
Metadata Entry (compact)
---
element: Customer.TaxNumber1
domain: Customer Master
system: SAP S/4
---
## Business metadata
- Definition: VAT or tax identification number for the customer
- Business owner: Finance Director
- Consumers: Billing, VAT reporting, EDI outbound
- Decisions made: Invoice tax calculation, compliance reporting
## Technical metadata
- Table: BUT000
- Field: STCD1
- Data type: CHAR(20)
- Source system: SAP MDG
- Update frequency: On change via MDG workflow
- Last modified: 2026-05-15
## Operational metadata
- Quality: 94% complete for DE customers; 62% for non-DE
- Validation: Required for DE; optional for others
- Known issues: CRM integration sometimes sends empty values
## Status
- Definition verified: Yes (by Finance Director, 2026-06-01)
- Owner assigned: Yes
- AI ready: Partial (quality gap noted)
Metadata Gap Note
| Element | Gap | Risk | Closure Action | Owner | Due Date |
|---------|-----|------|----------------|-------|----------|
| Product.CostGroup | No business definition | COGS report cannot be validated | Interview Product Manager | Data Steward | 2026-06-20 |
| SalesOrder.DeliveryBlock | Allowed values undocumented | Integration failures | Sample data and confirm with Logistics | Integration Lead | 2026-06-15 |
Quality checklist
- Every data element in scope has a business definition in plain language.
- Every element has a named owner who has confirmed the definition.
- Technical metadata (table, column, type) has been verified against the actual system.
- Allowed values are documented and confirmed.
- Consumers are listed and at least one has validated the usage description.
- All gaps are flagged with business risk and a closure action.
- The inventory has a maintenance owner and a review cycle.
Common mistakes
- Mistake: Cataloging only technical metadata (tables, columns, types) without business meaning. Consequence: The catalog is useless for business users, data scientists, and integration analysts.
- Mistake: Trusting outdated documentation without sampling the actual data. Consequence: The catalog documents a system that no longer exists; decisions based on it are wrong.
- Mistake: Letting technical teams define business meanings without business validation. Consequence: Definitions are technically accurate but business-wrong; reports and integrations fail.
- Mistake: Trying to catalog everything at once. Consequence: The effort never finishes; stakeholders lose confidence; the catalog is abandoned.
- Mistake: Producing a catalog with no maintenance plan. Consequence: The catalog is accurate on day one and stale by month three.
Agent instructions
When using this skill, an AI agent must:
- Scope tightly. Ask the user to name the specific system, domain, API, or report to catalog. Do not attempt enterprise-wide metadata extraction.
- Separate facts from assumptions. If a definition comes from documentation, label it "from docs." If it comes from expert interview, label it "from SME." If inferred from data, label it "inferred — needs validation."
- Produce structured entries. Output metadata entries in the compact template format. Do not produce narrative descriptions.
- Flag gaps explicitly. For every missing definition, owner, or consumer, produce a gap note with risk and action.
- Avoid inventing definitions. If the meaning of a field is unknown, say "unknown" and ask for an SME. Do not guess based on field names.
- Link to Atlas pages. If metadata gaps involve SAP master data, link to SAP Master Data Quality or Data Contracts.
- Handle missing information. If system access or SMEs are unavailable, produce a discovery checklist and ask the user to fill gaps.
Related skills
- Data Governance — when metadata gaps reveal ownership or rule problems.
- Data Lineage — when metadata must include source-to-consumer paths.
- Data Quality Root Cause — when metadata gaps cause defects.
- Master Data Management — when the metadata scope is master data.
Related Atlas pages
- Data Contracts — formal agreements that depend on accurate metadata.
- Data Quality Controls — controls that rely on metadata for enforcement.
- SAP Master Data Quality — metadata patterns for SAP master data.
- SAP Data Product — packaging data for consumption.
Verification status and limitations
This skill is a public working interpretation of metadata management practice. It is not official DAMA-DMBOK or ISO 11179 documentation. It has been applied in SAP and BI contexts but may need adaptation for cloud data lakes, NoSQL databases, or real-time streaming schemas.
Limitations: This skill does not cover automated metadata extraction tools, data catalog software selection, or semantic ontology design. It focuses on manual and semi-automated metadata inventory methods suitable for project and support contexts.