Data Management
This domain covers the disciplines used to acquire, store, move, and maintain data as an asset. The CDP exam tests whether you can distinguish modeling levels, apply integrity rules, choose integration patterns, and connect data quality and metadata to operational outcomes.
Core rules and concepts
- Model levels: conceptual models name business entities and relationships; logical models add attributes, keys, cardinality, and normalization; physical models add tables, indexes, partitions, and storage details.
- Relational integrity: a primary key is unique and non-null; a foreign key references a primary or unique key. Resolve many-to-many relationships with an associative entity carrying relationship attributes.
- Normalization: 1NF requires atomic values; 2NF removes partial dependencies on a composite key; 3NF removes transitive dependencies. Denormalize only for a documented performance or reporting reason.
- Transactions and keys: ACID protects concurrent transactions. Surrogate keys are system-generated; natural keys carry business meaning. Keep both when history and integration require them.
- Quality dimensions: accuracy, completeness, consistency, timeliness, uniqueness, and validity are measured against defined business rules, not assumed.
- Integration and metadata: ETL transforms before loading; ELT loads raw data and transforms in the target. Metadata includes technical structures and business definitions.
Worked example
A Student–Course many-to-many relationship needs an Enrollment associative entity. Enrollment holds student and course foreign keys plus grade and enrollment date. Grade belongs to the relationship, not to Student or Course. The composite key (student_id, course_id) enforces one enrollment per pair; a surrogate enrollment_id may be added for convenience. The design is in 3NF because non-key attributes depend on the key, the whole key, and nothing but the key.
Traps that cost marks
Confusing 2NF and 3NF; placing relationship attributes in the wrong entity; treating a surrogate key as the business identifier; ignoring optionality in cardinality; assuming denormalization is always wrong; and conflating master data with reference data.