Metadata management
Metadata management is learned and done in 6 activities, made of 90 knobits.
Learning path
- State the metadata requirements and strategy · 13 knobits
- Design the metamodel · 12 knobits · builds on State the metadata requirements and strategy
- Define business terms in the glossary · 18 knobits · builds on Design the metamodel
- Describe the catalogue entries · 12 knobits · builds on Design the metamodel, Define business terms in the glossary
- Trace column-level lineage · 11 knobits · builds on Describe the catalogue entries
- Analyse the impact of a change · 9 knobits · builds on Describe the catalogue entries, Trace column-level lineage
Foundation knowledge
15 knobits used across the activities rather than in one of them.
- M-901 Explains the business drivers of metadata management
- M-902 Explains the goals and principles of metadata management
- M-903 Explains metadata for unstructured data
- M-904 Explains the kinds of metadata sources
- M-905 Explains the types of metadata architecture
- M-906 Explains the steps of building a metadata strategy
- M-907 Explains the standards used for metadata
- M-908 Explains managing metadata stores
- M-909 Explains delivering metadata to its users
- M-910 Explains querying, reporting and analysing metadata
- M-911 Explains metadata repository management tools
- M-912 Explains metadata for big data ingest
- M-913 Explains readiness and risk assessment for metadata management
- M-914 Explains the organisational change metadata management needs
- M-915 Explains metadata management metrics
Which AI model is best at metadata management?
Best result for metadata management: gemini-4-argon (95% test score).
| Model | 1 | 2 | 3 | 4 | 5 | 6 | All |
|---|---|---|---|---|---|---|---|
| gemini-4-argon | 72% | 90% | 92% | 90% | 89% | 90% | 87% |
| gemini-3.1-pro-preview* | – | – | 81% | – | – | – | 81% |
| gpt-6-astra* | – | – | 94% | – | – | – | 94% |
| claude-fable-5-1* | – | – | 99% | – | – | – | 99% |
| gemini-3.8-flash* | – | – | 94% | – | – | – | 94% |
| gpt-6.1-sol* | – | – | 94% | – | – | – | 94% |
| claude-sonnet-5-5* | – | – | 98% | – | – | – | 98% |
| gemini-3.5-flash-lite | 0% | 14% | 46% | 48% | 73% | 80% | 44% |
| gpt-6-luna | 43% | 64% | 60% | 80% | 81% | 77% | 68% |
| claude-haiku-5-5 | 71% | 75% | 74% | 79% | 88% | 88% | 79% |
* From the earlier benchmark of 2026-10: 10 data-modelling test tasks, 3 runs per task, graded by claude-opus-5-5. These models have not taken the current 123 tasks yet; the two methods differ, so compare their figures with care. They are left out of the “most models” counts and the best-model picks.
Expected share of each activity's knobits the model masters; columns follow the learning path above. Method and caveats.