Visual Atlas — Maps 6–10
Map 6 — Dimensional Memory Picture
DIMENSION: Semester
│
DIMENSION: Campus ─── FACT TABLE ─── DIMENSION: Program
measurements
one declared GRAIN
│
DIMENSION: Student
Fact = measurements. Dimension = descriptive context. Grain = exactly what one fact row represents.
SCD: Type 1 = OVERWRITE
Type 2 = new ROW
Type 3 = new COLUMN
Conformed DIMENSION = shared descriptive meaning
Conformed FACT = shared measure definition
STAR = collapsed dimensions
SNOWFLAKE = normalized dimension hierarchies/components
Map 7 — Conceptual → Logical → Physical
CDM
high-level business concepts + relationships + vocabulary
↓
LDM
detailed technology-independent entities + attributes + domains + keys
↓
PDM
technology-specific tables + columns + datatypes + indexes + partitions + views
Separate path:
CANONICAL MODEL = shared message/payload structure for data in motion
Map 8 — Normalization Ladder
1NF → atomic values / no repeating groups / valid key / resolve M:M
↓
2NF → every non-key attribute depends on the COMPLETE minimal key
↓
3NF → no non-key dependency on another non-key attribute
↓
BCNF → stronger dependency rule
↓
4NF → multivalued dependencies
↓
5NF → join dependencies
Practice emphasis: 1NF–3NF. Unqualified “normalized model” normally means 3NF in Chapter 5.
Physical reminder: justified denormalization may later add redundancy; that does not make normalization useless.
Map 9 — Forward vs Reverse Engineering
FORWARD
Requirements → CDM → LDM → PDM
direction = increasing implementation detail
REVERSE
Existing Database → reconstructed PDM → inferred LDM → inferred CDM
direction = recovering abstraction / business meaning
Deciding clue: what evidence do you start with — requirements or implementation?
Map 10 — Model Governance and Quality
MODELING STANDARDS
↓ expected practice
MULTIDISCIPLINARY DESIGN REVIEW
↓ correctness / completeness / consistency
VERSION & CHANGE CONTROL
↓ Why | What/How | When | Who | Where
DATA MODEL SCORECARD
↓ measure quality across 10 dimensions
ONGOING MAINTENANCE
↺ keep model aligned with requirements + implementation
Scorecard 100-point memory
15 Requirements
15 Completeness
10 Scheme fit
15 Structural soundness
10 Generic structures
5 Naming
5 Readability
10 Definitions
5 Enterprise consistency
10 Metadata matches actual data
=100
Governance is not a one-time final approval.
Source anchors: current Visual Memory Map, Chapter 5 maps 6–10.