Data Management and Analytics
From raw data to trusted decisions
Data management covers governance, quality, metadata, master data, and the platforms that store and serve data. Candidates should distinguish a data warehouse (schema-on-write, curated, BI-oriented), a data lake (schema-on-read, raw and diverse), and a lakehouse (lake storage with warehouse-style transactions and cataloging). DAMA-DMBOK provides the vocabulary of data governance, stewardship, and quality dimensions: completeness, accuracy, consistency, timeliness, uniqueness, and validity. Know star schemas, fact grain, conformed dimensions, and slowly changing dimensions (SCD). ETL transforms before loading; ELT loads raw and transforms in the warehouse; CDC captures changes for near-real-time feeds.
Traps: calling a data lake a warehouse because both store data; using data lake as an excuse to skip governance-an unmanaged lake becomes a swamp; and mixing SCD types. SCD Type 1 overwrites history; Type 2 adds a new row with effective dates and a current flag; Type 3 adds a previous-value column. Another trap is stating a metric without its grain: revenue by day differs from revenue by customer by day.
Worked example: a customer changes address. Analysts need historical region reporting, so use SCD Type 2: close the old row (end date, current flag = N), insert a new row with a new surrogate key, and preserve the natural or business key. Fact rows keep the surrogate key that was current when the event occurred.
Rule of thumb: declare the grain before choosing the model; govern the definition of every shared metric once, centrally.