Integrated Capstone — University Modernization
Situation
A university modernization program has:
- inconsistent definitions of Student and Course;
- a legacy relational database with undocumented relationships;
- a new analytics requirement for Enrollment measures by Semester and Campus;
- duplicated reference-code tables;
- model files stored on individual laptops;
- delivery teams that alter physical tables without updating logical models.
Leadership asks the modeling team to “create the database design.”
Strong Chapter 5 response
A physical schema is only one downstream deliverable. The modeling team should:
- Establish scope and terminology with a CDM. Resolve Student, Course, Enrollment and related business vocabulary with business SMEs/stewards.
- Build/refresh the LDM. Model entities, relationships/cardinality, attributes, domains, keys, normalization, and definitions while remaining technology-independent.
- Model analytics dimensionally. Define Enrollment business process, declare grain, identify facts/measures and Semester/Campus dimensions, and decide history/conformance needs.
- Reverse engineer the legacy implementation. Recover PDM → LDM → conceptual understanding and compare it to the desired model.
- Capture lineage/source-target mappings. Connect model attributes to real operational sources and transformations.
- Resolve duplicated reference structures deliberately. Determine authoritative definitions/domains rather than copying the current duplication forward by accident.
- Create the PDM intentionally. Resolve subtypes, physical names/datatypes/reference data, key implementation, indexes, partitions, views/materialized views, and denormalization only where requirements justify them.
- Move model knowledge into governed shared Metadata. Do not leave the authoritative model on individual laptops.
- Institute standards and multidisciplinary reviews. Test correctness, completeness, consistency, scheme fit and implementation viability.
- Use version/change control and ongoing maintenance. Physical implementation changes must be reconciled back to the governed model.
- Use the Data Model Scorecard. A model is not high quality merely because it is readable.
Industry-model question
Should the team automatically replace its models with an industry data model?
No. Industry models are broad pre-built starting/reference content. They may accelerate discovery but must be customized to the organization’s actual requirements.
Capstone teach-back
Explain why each of these alone is insufficient:
- “just draw the PDM”;
- “just normalize everything”;
- “just buy an industry model”;
- “just use a modeling tool”;
- “just approve the model once.”
A strong answer should connect requirements, model levels, scheme choice, lineage, physical trade-offs, governance, quality, and maintenance.