Purpose

Identify data entities, relationships and attributes including the appropriate characteristics without regard for their application usage. Optionally, create a corporate or application-spanning entity/relationship model.

Benefits:

• Encourage subject matter expert interaction with the business data model
• Identify business data entities and attributes before design
• Resolve data issues with the appropriate level of authority
• Illustrate the constraints that data place on business decision-makers
• Ensure fit with corporate data strategy
• Facilitate evolution to data marts and warehouses
• Reduce the potential for expensive changes later in the project

Focus

Identify and define the data needed to support the business. This data does not include data needed “because the database software or organizational structure, or legal constraints, or software package requires it. This is not the time to worry about data elements specific to the physical implementation of a given solution.

You will need:

Inputs

Existing reports, forms, screens, data structures, etc. Existing corporate or related data models

Contributors

Customer representatives, Systems analysts, Business analysts

Experts

Data administrator; Data analysts

Advisors

Data base administration

Techniques

Interviewing techniques, Data modeling, Data normalization

Tools

Data or Object modeling, Data dictionary, Word processor

Agenda

The following example depicts the sequence of activities during a typical 5 day session.

Time

Participant activities

Day 1, a.m.

• Introductions and agenda presentation

• Identify first level data entities

Day 1, p.m.

• Create intuitive data model

• Define key attributes for entities

Day 2, a.m.

• Expand key attribute metadata

• Select base data elements in existing outputs

Day 2, p.m.

• Resolve derivable data elements

• Create expanded attribute definitions

Day 3, a.m.

• Normalize data from existing outputs

• Expand on attribute metadata

Day 3, p.m.

• Evaluate data attribute orphan list and assign to entities

• Initiate data attribute ownership determination

Day 4, a.m.

• Combine intuitive and normalized data models

• Resolve model discrepancies

Day 4, p.m.

• Collect additional attribute definitions and metadata

• Finalize data attribute ownership assignments

Day 5, a.m.

• Finalize combined data model

• Validate data attribute synonym definitions

Day 5, p.m.

• Revisit open issues and questions lists

• Assign post-session tasks and schedule post-session meeting

Tips and Tricks

A business data model (or entity/relationship diagram,) is relatively stable. Changes reflect business decisions in the real world, which is what the data represents. Physical data structures may change often.. Keep your logical model clean.

A one-week session for the creation of a preliminary entity/relationship diagram (ERD) suffices for a very limited scope. As a rule-of-thumb, the delivered model contains somewhere between 50 and 100 entities with the associated attributes and relationships. Depending on the size and complexity of the business involved, an enterprise-wide ERD can easily require 4 - 8 one-week sessions. In this case, plan a final, “consensus-building” session to iron out discrepancies between individual session deliverables.

Output description

The main tangible deliverable of this session is a data model. A major intangible deliverable is recognition of the limits that the current data structures place on the business’s ability to react to changing conditions.

Detailed Entity-Relationship Diagrams depict the objects (entities) that describe the business world, the relationships between objects and their primary key attribute(s).

Attributes are the data elements that describe the objects and/or allow for unique identification of an object.

Attribute characteristics is data about data. A data dictionary is a repository for:

Primary name is the official designator for the attribute.

Synonyms describe what other names different organizational units call this attribute.

Definitions are a single, simple English sentence expressing what information the attribute contains.

Data rules constrain the structure of information stored in the attribute, e.g., numeric/alpha, number of characters, storage type, etc.

Validity rules constrict the values stored in the attribute.

Default values are assigned in the absence of an explicit value.