Data Governance
11%of exam
StewardshipGovernance CouncilOperating ModelsDAMA Definitions
Data Modeling and Design
11%of exam
Data Quality
11%of exam
DQ DimensionsDMAICRoot CauseProfiling
Metadata Management
11%of exam
Master and Reference Data
10%of exam
Data Warehousing and BI
10%of exam
Data Architecture
6%of exam
Zachman FrameworkEnterprise ModelData FlowTOGAF Alignment
Document and Content Management
6%of exam
ECMRetentione-DiscoveryGARP
Data Integration and Interoperability
6%of exam
Data Security
6%of exam
CIA TriadRBACEncryptionPrivacy Law
Data Storage and Operations
6%of exam
DBA RoleNoSQLBackup RecoveryPerformance Tuning
Big Data and Data Science
2%of exam
Four VsData LakeHadoop SparkAnalytics Maturity
Data Ethics
2%of exam
BiasFairnessTransparencyBelmont Principles
Data Management Process and Maturity
2%of exam
Environmental HexagonCMMI-DMMDCAMMaturity Assessment
Quick Facts
- Exam
- CDMP Fundamentals
- Body
- DAMA International
- Questions
- 100
- Time
- 90 min (110 ESL)
- Pass
- 60/70/80%
- Format
- Open-book, online-proctored
- Level
- Associate/Practitioner/Master
- Blueprint
- DMBoK2 Revised (Mar 2024)
Steward vs Custodian
Data Steward
- Owns business meaning
- Approves definitions and rules
Data Custodian
- Operates data systems
- Executes technical storage tasks
Meaning vs operations
Governance Operating Model Picker
- One team decides everything→Centralized
- Each unit decides alone→Decentralized
- Local stewards, central council→Federated
- Mix of central and local→Hybrid
Governance Roles
- Data Steward
- Owns business meaning
- Data Custodian
- Operates data systems
- Data Governance Council
- Approves policy, standards
- Data Mgmt Executive
- Leads governance program
- Business Glossary
- Enterprise term definitions
- Data Owner
- Accountable for asset
Governance vs Management
Governance
- Authority and control
- Sets policy direction
Management
- Executes plans and practices
- Runs daily operations
Authority vs execution
Governance Operating Models
- Centralized
- Single team decides
- Decentralized
- Business units decide
- Federated
- Local stewards, central council
- Governance
- Authority and control
- Management
- Execution of plans
Model Levels
- Conceptual Model
- Business terms only
- Logical Model
- Attributes, keys, tech-neutral
- Physical Model
- Platform-specific structures
- Enterprise Data Model
- Subject areas enterprise-wide
Normalization and Notations
- 1NF
- Atomic values only
- 2NF
- No partial dependency
- 3NF
- No transitive dependency
- BCNF
- Tightens 3NF, multiple keys
- IE Notation
- Crow's-foot relationships
- Naming Standards
- Class word plus modifier
DQ Dimensions Mnemonic
CACTUV covers six DQ dimensions
C: Completeness presentA: Accuracy correctC: Consistency agreesT: Timeliness freshU: Uniqueness no dupesV: Validity conforms
Which DQ Dimension Failed
- Value is missing→Completeness
- Value is wrong→Accuracy
- Systems disagree→Consistency
- Data is stale→Timeliness
- Records are duplicated→Uniqueness
- Value breaks a rule→Validity
Data Quality Dimensions
- Completeness
- Required values present
- Accuracy
- Matches real-world truth
- Consistency
- Systems agree
- Timeliness
- Data fresh enough
- Uniqueness
- No unintended duplicates
- Validity
- Conforms to rules
DQ Improvement Process
- Data Profiling
- Discovers actual characteristics
- Root Cause Analysis
- Finds underlying source
- DMAIC
- Define Measure Analyze Improve Control
- PDCA
- Plan Do Check Act
- Cost of Poor Quality
- Rework, churn, fines
- DQ Scorecard
- Measurable rule conformance
Business vs Technical Metadata
Business Metadata
- Ownership and definitions
- Policies and business rules
Technical Metadata
- Column data types
- Table and object structure
Meaning vs structure
Metadata Types
- Business Metadata
- Ownership, definitions, rules
- Technical Metadata
- Structures, data types
- Operational Metadata
- Load times, job status
- Data Lineage
- Source-to-consumption traceability
Metadata Repository Architecture
- Centralized Repository
- One physical store
- Distributed Repository
- Metadata stays at source
- Hybrid Repository
- Combines both patterns
- CWM
- OMG DW/BI metadata standard
MDM Architecture Spectrum
Registry to Consolidation to Coexistence to Transactional
Registry: pointers onlyConsolidation: analytics copy onlyCoexistence: syncs both waysTransactional: hub is authoritative
Registry vs Transactional MDM
Registry
- Pointers only
- No source changes
Transactional
- Hub is authoritative
- Writes back to sources
Lightest touch vs heaviest
MDM Architecture Picker
- No source changes allowed→Registry(Pointers only)
- Need analytics copy only→Consolidation(No writeback)
- Need sync both directions→Coexistence(Persists and syncs)
- Hub is system-of-record→Transactional(Writes back to source)
MDM Architectures
- Registry
- Pointers only, no copy
- Consolidation
- Analytics copy, no writeback
- Coexistence
- Persists and syncs back
- Transactional
- Hub is system-of-record
Master Data vs Reference Data
Master Data
- Core entities like Customer
- Slowly changing over time
Reference Data
- Small code lists
- Constrains operational values
Core entity vs code list
MDM Core Concepts
- Golden Record
- Trusted consolidated entity
- Match-Merge
- Resolves duplicate identity
- Survivorship
- Rules pick surviving values
- Cross-Reference (xref)
- Maps IDs to master
- Reference Data
- Small authoritative code lists
Warehouse Architecture Camps
Inmon normalizes first; Kimball dimensions first
Inmon: enterprise warehouse firstKimball: dimensional marts firstODS: near-real-time operational layerMart: subject-area subset only
Inmon vs Kimball
Inmon CIF
- Normalized warehouse first
- Feeds marts downstream
Kimball Bus
- Dimensional marts first
- Conformed dimensions integrate
Top-down vs bottom-up
Warehouse Architecture Picker
- Need normalized enterprise warehouse→Inmon CIF
- Need fast dimensional delivery→Kimball Bus
- Need near-real-time reporting→ODS
- Need one subject-area subset→Data Mart
Warehouse Architectures
- Inmon CIF
- Normalized enterprise warehouse first
- Kimball Bus
- Dimensional marts, conformed dims
- ODS
- Near-real-time operational layer
- Data Mart
- Subject-area subset
Star vs Snowflake Schema
Star Schema
- Denormalized dimensions
- Fewer joins, faster
Snowflake Schema
- Normalized dimensions
- More joins, less redundancy
Speed vs storage efficiency
Dimensional Modeling
- Star Schema
- Fact center, denormalized dims
- Snowflake Schema
- Normalized dimension tables
- Conformed Dimension
- Shared across fact tables
- SCD Type 1
- Overwrite, no history
- SCD Type 2
- New row, tracks history
- Atomic Grain
- Lowest-level fact detail
BI Delivery
- OLAP
- Drill-down, slice-and-dice
- ELT
- Transform inside warehouse
- ETL
- Transform before loading
- Self-Service BI
- Certified datasets, curated marts
Architecture Frameworks
- Zachman Framework
- Perspectives x aspects grid
- TOGAF ADM
- Enterprise architecture method
- Data Flow Diagram
- Movement between systems
- Enterprise Data Architecture
- Subject areas, flows, stores
ECM and Records
- ECM
- Enterprise content management
- Retention Schedule
- Defines how long kept
- e-Discovery
- Legal content search
- GARP Principles
- ARMA records management principles
- Taxonomy
- Classifies unstructured content
ETL vs ELT
ETL
- Transform before load
- Separate transform tier
ELT
- Transform inside warehouse
- Uses warehouse compute
Where transformation happens
Integration Patterns
- ETL
- Extract, transform, load
- CDC
- Captures source changes
- ESB
- Message-based integration bus
- Canonical Model
- Common integration message format
- Virtualization
- Query without moving data
Security Fundamentals
- CIA Triad
- Confidentiality, integrity, availability
- RBAC
- Role-based access control
- ABAC
- Attribute-based access control
- Encryption
- Protects data confidentiality
Privacy Regulations
- GDPR
- EU data privacy law
- HIPAA
- US health data privacy
- CCPA
- California consumer privacy act
- Data Classification
- Labels sensitivity level
Storage and Operations
- DBA Role
- Operates database platforms
- NoSQL
- Non-relational database
- Backup and Recovery
- Restores after failure
- Performance Tuning
- Optimizes query speed
Big Data Concepts
- Volume
- Data size scale
- Velocity
- Speed of arrival
- Variety
- Structured, unstructured mix
- Veracity
- Trustworthiness of data
- Data Lake
- Raw, schema-on-read storage
Data Ethics Concepts
- Bias
- Systematic unfair skew
- Fairness
- Equitable treatment of subjects
- Transparency
- Explainable data use
- Belmont Principles
- Respect, beneficence, justice
Environmental Factors Hexagon
Goals Activities Roles Deliverables Practices Tools
Goals: why it mattersActivities: what gets doneRoles: who does workDeliverables: outputs producedPractices: how it's doneTools: technology used
Environmental Factors and Maturity
- Goals and Principles
- Why the work matters
- Activities
- What work gets done
- Deliverables
- Outputs of the work
- Roles and Responsibilities
- Who performs the work
- CMMI-DMM
- Data maturity model
- EDM Council DCAM
- Capability assessment model
Common Traps
Steward ≠ Custodian
Steward owns business meaning ≠ Custodian operates the systems
Governance ≠ Management
Governance sets authority, policy ≠ Management executes plans, practices
Master Data ≠ Reference Data
Master: core changing entities ≠ Reference: small static code lists
Star Schema ≠ Snowflake Schema
Star denormalizes dimension tables ≠ Snowflake normalizes dimension tables
ETL ≠ ELT
ETL transforms before loading ≠ ELT transforms after loading
Data Quality ≠ Data Profiling
Quality measures rule conformance ≠ Profiling discovers actual characteristics
Policy ≠ Standard
Policy states business intent ≠ Standard sets measurable, testable rules
Golden Record ≠ Source Record
Golden record is trusted, consolidated ≠ Source record is raw, unmerged
Last Minute
- 1.Four core areas: 11% each
- 2.MRD and DW/BI: 10% each
- 3.Five areas share 6% weight
- 4.Three areas share 2% weight
- 5.Steward owns meaning, not systems
- 6.Custodian operates systems, not meaning
- 7.MDM order: Registry Consolidation Coexistence Transactional
- 8.Inmon normalizes; Kimball builds dimensionally
- 9.Star denormalized; snowflake normalizes dimensions
- 10.SCD Type 2 adds history rows
- 11.Pass scores: 60/70/80 percent
- 12.100 questions, 90 minutes, open-book
Buy on Amazon
Take courses on Udemy