Skills in action / Business catalog
INDEPENDENT CONSULTANT · DATA GOVERNANCE, DATA QUALITY & MDM
Training and coaching for Data Owners and Data Stewards, built on a real case
Most organisations already have plenty of metadata: table names, column lists, data dictionaries. What they miss is a catalog that a business person can read. I teach Data Owners and Data Stewards how to build it: which terms deserve a place, how to organise them, how to share common concepts between domains, and how to keep the result trustworthy over time. The training does not start from theory, but from a real catalog of open data that we structure together, step by step.
Terms are the nouns the business uses: Municipality, never CD_MUNTY_REFNIS.
A concept is defined once; the other names become synonyms.
Person, Address or Country are defined once for the whole organisation.
No deeper than the level at which people still ask questions.
Only reviewed definitions reach the users, with a named owner.
The Belgian statistical office, Statbel, publishes its open data with a standard catalog file (DCAT). It is well documented technically, but it has no business structure: it tells you which files exist, not what the data means. That makes it the perfect training ground: real data, real ambiguity, and nothing confidential.
| In the source | What it hides | In the business catalog |
|---|---|---|
CD_SEX, CD_GENDER, MS_SEX_ … 9 spellings | One concept, many names and languages | One term: Sex, with synonyms in French and Dutch |
| Territorial codes under 113 column names | The same code list used for residence, birth, work, accidents… | One shared hierarchy (Region › Province › Municipality) and one role per context |
TYPE_ENT: natural or legal person | The split between consumers and businesses | The generic Party: a Person or an Organisation, and their relationships |
MS_DEAD, MS_SLY_INJ, MS_VIC_OK … | A classification spread over six columns | Number of Victims by Injury Severity |
Source data: Statbel open data (statbel.fgov.be/fr/open-data), reused under CC BY 4.0.
A business catalog needs only a few levels. Each one answers a different question and has an owner. The technical layer (datasets and columns) is harvested by the tool and linked to the business terms; nobody types it by hand.
CD_REFNIS_LVL_4The type tells the reader how to use a term, and the steward which naming rule applies.
Groups terms to navigate. Vital Events
A thing with its own identity. Person, Dwelling
Describes an object. Sex, Legal Form
A shared concept seen from one context. Place of Birth
A number the business reads. Population Count
An official code list. Economic Activity (NACE-BEL)
Identifies one instance. Enterprise Number
The method meets in the middle: the data gives the evidence of what is really used, the business gives the structure. Reverse engineering alone produces a list of columns with nicer names; interviews alone produce terms that no data supports.
Concepts used by several domains are placed in one shared glossary owned by the governance office. Domains link to them instead of redefining them, so that data from different domains can be combined later, for example a consumer view and a business view of the same person.
Defined once, reused by every domain
Participants leave with three things that work together: a rule book, a method and a catalog that the business can already use.
Who decides, and what must a term contain?
Which terms, and how are they organised?
Can a business user find and trust a definition?
Search a word in English, French or Dutch, browse by domain, and open a term: definition, synonyms, related terms and owner. Only approved definitions are shown. The page is generated from the catalog tool, so it stays in line with the governed content.
Decide the scope of their glossary, what they approve and which shared concepts they must not redefine.
A repeatable method to go from datasets and columns to well-placed, well-defined terms.
How columns are profiled and linked to terms, and why codes and labels are not terms.
How to run the shared vocabulary, the approval workflow and the publication to business users.
The workshop alternates short explanations and exercises: reading a mountain of data, stripping the technical noise, deciding "term or not", modelling roles, writing definitions, importing and approving terms in the catalog tool. The coaching then applies the same method to your domains, in French, English or Dutch.
Whether you are starting a catalog or trying to make an existing one useful, I adapt the training to your tool (OpenMetadata, Collibra, Purview or another catalog) and to your domains, and I leave your teams able to continue on their own.
Contact me Open the demo My four-phase approach The technical note: the Statbel pipeline (EN)