Property and Casualty Insurance Working Group
Logical Relational Data Modeling Standards
Versio n 1.0 Property and Casualty Insurance Working Group
Jun e 16, 2008
Table of Contents Introduction ...... 4
Purpose ...... 4
Document Maintenance ...... 4
Scope ...... 4
Logical Relational Data Model Definition ...... 4
ER Diagramming Conventions ...... 6
Logical Relational Data Modeling Standard Page 2 Property and Casualty Insurance Working Group
Modeling Syntax ...... 6
Diagramming Layout Guidelines ...... 7
Normal Forms ...... 8
Writing Definitions of Logical Objects ...... 9
Logical Object Definition Guidelines: ...... 9
Entity Definition Guidelines: ...... 9
Attribute Definition Guidelines : ...... 9
Naming Logical Objects ...... 10
Logical Object Naming Guidelines ...... 10
Entity Naming Guidelines ...... 11
Attribute Naming Guidelines ...... 11
Relationship Naming Guidelines ...... 13
Relationship Standards ...... 14
Super- types and Sub- types ...... 15
Entity Keys ...... 17
Dimensional Data Modeling ...... 18
Appendix ...... 19
Class Words ...... 19
Logical Relational Data Modeling Standard Page 3 Property and Casualty Insurance Working Group
Introduction Purpose This document provides standards and guidance for the naming and use of objects in logical relational data models. Logical objects are created and maintained to meet business requirements. Accurate naming clarifies the specific nature of each logical object. Consistency allows the logical names to have persistent value in differentiating data items. Name formation and the use of logical modeling objects are independent of any particular data modeling tool or relational database management system (RDBMS) platform. These logical relational data modeling guidelines are independent of specific CASE tools.
The intention of this standard is to establish an agreed- upon basis for developing logical relational data models in order to promote greater quality and consistency across data models and enable objective model reviews.
Document Maintenance To suggest improvements, changes or additions to this standard, contact:
Gail Austin or Harsh Sharma [email protected] [email protected]
Scope These standards apply to all logical relational data models that are developed by OMG submission teams.
Logical Relational Data Model Definition The relational model for database management is a database model based on predicate logic and set theory. It was first formulated and proposed in 1969 by Edgar Codd with aims that included avoiding, without loss of completeness, the need to write computer programs to express database queries and enforce database integrity constraints. “Relation” is a mathematical term for “table”,
Logical Relational Data Modeling Standard Page 4 Property and Casualty Insurance Working Group and thus “relations” roughly means “based on tables”. It does not refer to the links or “keys” between tables, contrary to popular belief.1
A logical relational data model defines what an organization knows about things of interest to the business and graphically shows how they relate to each other in an entity relationship (ER) diagram. An entity relationship diagram is an abstract conceptual representation of structured data. It uses standard symbols to denote the things of interest to the business (entities), the relationships between entities and the cardinality and optionality of those relationships. The
Logical Relational Data Model, in contrast to the more abstract Conceptual Relational Data Model, contains detailed characteristics of the entities (attributes) and their definitions. It generates the structure of a physical data model which in turn generates a database following Model Driven
Architecture principles. It is a result of detailed analysis of the business requirements.
The following illustration shows how the logical model fits into the overall data modeling process:
1 Wikipedia – relational model
Logical Relational Data Modeling Standard Page 5 Property and Casualty Insurance Working Group
Ultimately, the logical relational data model helps to solidify and validate business requirements and delivers stable, flexible data structures that are easily navigated and can answer unanticipated questions.
ER Diagramming Conventions Modeling Syntax The recommended notation for models is Information Engineering (IE) – “Crow’s Feet” - because it is easier for users to interpret than the Integration Definition for Information Modeling (IDEF1X) notation.2
2 The choice of IE notation will be revisited when the Barker notation becomes more widely available in the modeling tools.
Logical Relational Data Modeling Standard Page 6 Property and Casualty Insurance Working Group
Diagramming Layout Guidelines Orient entities so that the “toes” of a relationship’s crow’s foot always point down. This puts fundamental entities in the top area of the diagram, and positions associative and subtype entities in the lower area of the diagram.
Recommended crow’s feet down convention Avoid dead crows!
CONTACT PROFILE PERSON Person Identifier (FK) Person Identifier (FK) Contact Point Identifier (FK) First Name Home Contact Point Indicator Middle Name Work Contact Point Indicator Last Name Legal Name Nickname CONTACT POINT Name Suffix Contact Point Identifier Birth Date Birth Place Name Gender Code
PERSON CONTACT POINT Person Identifier (FK) Contact Point Identifier First Name Middle Name Last Name CONTACT PROFILE Legal Name Nickname Person Identifier (FK) Name Suffix Contact Point Identifier (FK) Birth Date Home Contact Point Indicator Birth Place Name Work Contact Point Indicator Gender Code
Keep the relationship lines as straight as possible. Avoid unnecessary bends. Too many symbols clutter the diagram and make it confusing to the viewer.
Avoid crossing relationship lines. Crossed lines make it difficult to understand which entities are related.
Relationship names should be placed on the diagram so that the verbs or verb phrases are read in a clockwise direction from one entity to the related entity.
Example:
Logical Relational Data Modeling Standard Page 7 Property and Casualty Insurance Working Group
POLICY Policy Identifier
covers
is covered by
EXPOSURE Policy Identifier (FK) Insured Object Identifier (FK) Coverage Type Identifier (FK)
Normal Forms Normal Forms provide a way to structure data to eliminate undesirable redundancies, inconsistencies and dependencies. Normalization is a formalized technique for creating the most desirable logical model for the given data and business rules. Completed logical models should be in, at least, Boyce/Codd Normal Form (BCNF)3. For a model to be in BCNF, every entity in the model must be in BCNF. The normal forms are summarized below:
Firs t Nor ma l For m (1NF) identifies and eliminates repeating groups and establishes a primary key.
Secon d Nor ma l For m (2NF) identifies and removes partial- key dependencies. This applies
only to tables with composite keys.
Third Nor ma l For m (3NF) identifies and eliminates non- key attributes that are dependent
on other non- key attributes.
Boyce/Cod d Nor ma l For m (BCNF) identifies and eliminates key attributes that are
dependent upon other key attributes in an entity with a composite key.
3 See Wikipedia Database Normalization: http://en.wikipedia.org/wiki/Database_normalization
Logical Relational Data Modeling Standard Page 8 Property and Casualty Insurance Working Group
Writing Definitions of Logical Objects Good Logical Object names are important because they provide a persistent record of the unique nature of each object. Good names cannot be developed unless the object first has a good business definition.
Logical Object Definition Guidelines: Use industry definitions where possible and appropriate. Describe what the entity or attribute is – not where, when or by whom it is used. Be clear and concise. Write as if the reader is unfamiliar with the business area. Use business terms rather than technical terms to express the meaning and importance to the business. Use mixed case according to standard business English conventions. Do not use jargon, abbreviations or acronyms. Do not include information that should be documented elsewhere, such as process descriptions.
Entity Definition Guidelines: Entity definitions should be robust and communicate the essential and unique business nature of the entity. Do not depend on or refer to the definition of another object in the model. Express one concept or idea – each entity should have a unique meaning. Attribute Definition Guidelines: Attribute definitions should communicate the essential business nature and purpose of the attribute. Do not depend on or refer to the definition of another object in the model, except for derived attributes. Include the domain of allowed values and default value where appropriate.
Logical Relational Data Modeling Standard Page 9 Property and Casualty Insurance Working Group
Naming Logical Objects Logical Object Naming Guidelines Use one or more words which are formed using the 26 letters (A- Z), the 10 digits (0- 9), and no special characters. Separate words in the name with one space Spell out words completely using no abbreviations. Use the minimum set of words for the name that completely and uniquely capture the concept expressed in the business definition Reflect the business nature of the object in its name Review names and corresponding definitions with business subject matter experts and get their approval Express a single idea or concept in the name that is clear and self- explanatory. Write in plain English, spelling out all terms in full using business terms as defined by the business client or as defined in a business or industry dictionary. Do not use the possessive form; the articles “a”, “an”, or “the”; conjunctions; verbs; or prepositions in the name. Do not use the names of organizations, departments, computer applications, reports, publications, forms or computer screens in the name. Exceptions Acronyms – An acronym is a word formed from the initial letters of a name, as WAC for Women’s A rmy Corps, or by combining initial letters or parts of a series of words, as rada r for ra dio detecting and r anging. When an acronym is widely known it may be an exception to the no abbreviation rule. A list of exceptions should be maintained as an appendix to this standard and subject to an approval and a governance process. Abbreviations – if the object name is too long to fit in the space allotted by the data modeling tool and all non- essential words have been eliminated from the name, abbreviate the class word. If the name is still too long, find text in the name that can form acronyms. Starting with the right- most text, apply the acronym and repeat moving left in the name until the name fits. Hyphen – use if the correct spelling of the word contains a hyphen (e.g. off- premises) Slash – allowed if used in a business term (e.g. Actual/Expected)
Logical Relational Data Modeling Standard Page 10 Property and Casualty Insurance Working Group
Camel Case – allowed if the business term has an uppercase letter beyond the first letter – though rarely found in formal written English, it is sometimes found in product names or company names (e.g. NetQuote, SmartBrief)
Entity Naming Guidelines Form a meaningful, concise, descriptive business name for the entity by extracting the important concepts from its business definition. The name should avoid confusion with
similarly named but differently defined entities in other business areas.
Use business terms as defined by a business subject matter expert or by a business dictionary.
Make the entity name a singular noun or noun phrase with qualifying adjectives because each instantiation of the object represented by the entity is a single thing.
Use UPPER CASE.
Consider appending “LOOKUP” to reference entity names to make them easier to distinguish from fundamental entities.
Do not use the words “Entity” or “Table” in the entity name unless they are part of common business terminology.
Combine the names of the parent entities to form the name of the associative entity if that forms a meaningful business name. For example, PERSON SKILL describes the association
between the PERSON and SKILL entities. In other cases, the noun form of the relationship
verb may form the associative entity name as in POLICY describes the association between
PARTY and PARTY.
Attribute Naming Guidelines Form a meaningful, concise, descriptive business name for the attribute by extracting the important concepts from its business definition. Attributes in more than one model should
Logical Relational Data Modeling Standard Page 11 Property and Casualty Insurance Working Group
have the same name and definition in all models.
Use a singular noun or singular noun phrase with qualifying adjectives that are meaningful to the business.
Use Title Case.
Do not use a class word or its abbreviation by itself as an attribute name.
Do not use the word “Attribute” in the attribute name unless it is part of common business terminology.
Attribute Name Structure
o An attribute name begins with at least one Qualifier followed by a Class Word. Note that conjunctions, verbs and other parts of speech are eliminated when they do not
affect the meaning of the name.
o Class words describe the type of data identified by the attribute name. Examples include: amount, code, date, indicator, name and number.
o End the name with an approved class word that best categorizes the attribute. Class words may also give an indication of the data type and possible values of the
attribute, e.g. an indicator is always a single alphanumeric character with only 2
possible values other than Null, ‘Y’ or ‘N’.4
o Units of Measure describe the quantity that was measured such as height or volume.
o Objects are used for program objects, images, sounds and videos.
4 See Appendix for details on Class Words.
Logical Relational Data Modeling Standard Page 12 Property and Casualty Insurance Working Group
Examples of logical attribute names and their components:
QUALIFIERS CLASS WORDS
MODIFIER PRIME WORD KEY WORD UNIT OF MEASURE OBJECT
Automobile Acquisition Date
Insurance Company Name
Payment Status Code
Valid Driver License Indicator
Vehicle Engine Capacity Cubic Centimeters
Accident Photograph Image Jpeg
Relationship Naming Guidelines The relationship name should be a verb or a verb phrase in third person singular form, i.e. a verb form that is appropriate for a singular occurrence of the entity. This verb or verb
phrase should be an active verb in the parent to child direction and a passive verb phrase
in the child to parent direction. When used with the cardinality and optionality information,
the verb or verb phrase allows the relationship to be read as bi- directional English
sentences. For example: A POLICY covers zero, one or many EXPOSURE(S). An
EXPOSURE is covered by one and only one POLICY.
Logical Relational Data Modeling Standard Page 13 Property and Casualty Insurance Working Group
Do not include words that convey cardinality or optionality in the verb phrase – words such as ‘may’, ‘must’, ‘one and only one’ or ‘one or many’ are derived from the relationship
symbols.
Avoid using generic or vague words and phrases such as ‘is’, ‘has’, ‘consists of’, ‘relates to’, ‘associated with ‘, etc.
Relationship Standards A relationship describes the precise business rules governing the association between two entities and facilitates the identification of foreign keys and referential integrity rules that may be required in the database design.
The minimum components that must be specified for each relationship are:
o Name – a verb or verb phrase from parent to child
o Optionality rules
o Cardinality rules
o Qualification as an identifying or non- identifying relationship
Many- to- many relationships are desirable in Conceptual Data Models but should always be resolved with an associative entity in a Logical Data Model even if the associative entity
has no attributes other than the keys.
Investigate all mandatory one- to- one relationships because usually the two entities are in fact one entity.
Eliminate circular relationships because they cause problems establishing proper data dependency sequences. They usually result from an incorrect or misunderstood business
rule.
Logical Relational Data Modeling Standard Page 14 Property and Casualty Insurance Working Group
Eliminate redundant relationships that consist of two dependency paths between the same two entities. One of the paths is a direct relationship between the entities; the other uses a
non- direct path that involves other entities. These redundant relationships may lead to
problems with database consistency.
Carefully review multiple relationships between the same two entities as they tend to represent process logic and may introduce conflicting cardinalities. If the multiple
relationships are created to document roles, a better solution may be to create a role entity
with appropriate subtypes.
Super-types and Sub-types Super- types and sub- types can be the result of either a generalization process – bottom- up – or a specialization process – top- down. The result is a super- type (parent) that contains attributes that are shared by all subtypes and a sub- type (child) that inherits all the shared attributes from the super- type but also has unique attributes of its own.
A sub- type has an ‘is a’ relationship to its super- type. Sub- types are not ‘composed of’ relationships.
Super- types and sub- types clarify complex business rules and constraints between entities.
The super- type and sub- type have an exclusive OR relationship. An instance of the super- type can be an instance of only one of the sub- type entities.
Logical Relational Data Modeling Standard Page 15 Property and Casualty Insurance Working Group
An example of super- types and sub- types:
INSURED OBJECT Insured Object Identifier
Geographic Location ID (FK)
VEHICLE HOME Insured Object Identifier (FK) Insured Object Identifier (FK) Registration State Code (FK)
MOTORCYCLE RECREATIONAL VEHICLE Insured Object Identifier (FK) Insured Object Identifier (FK)
AUTOMOBILE Insured Object Identifier (FK)
Logical Relational Data Modeling Standard Page 16 Property and Casualty Insurance Working Group
Entity Keys A key identifies specific occurrences of an entity. They can be simple, consisting of a single attribute, or they can be composite, consisting of two or more attributes.
A Candida te Key uniquely identifies occurrences of an entity. There may be more than one candidate key for an entity. Candidate keys are not usually recorded in the logical
data model because they become either a primary key or an alternate key.
A Primary Key is a single candidate key selected as the ‘primary’ unique identifier for the entity.
o The primary key must be stabl e for a relational data model. If the value were to change over time, the result could be either a non- unique key value or multiple key
values for one instance of an entity. Either situation could cause ambiguous or lost
data, system crashes or difficult update processes.
o The primary key should be definitiv e because it uniquely identifies an instance of the entity and thus no instance can be added to the entity until its identity is fully
known. The primary key cannot be nullable or contain nullable components.
o The primary key should use the minima l number of attributes required to define a unique instance of the entity. A concise key has advantages in the physical
database such as smaller indexes and foreign keys.
An Al terna te Key is any candidate key not selected as the primary key of an entity. Alternate keys are not usually recorded in the logical model but may become indexes in
the physical model. Alternate keys are usually unique but are not required to be.
A Surroga te Key consists of a single attribute created for the sole purpose of uniquely identifying an instance of an entity. Natural keys consist of attributes that ‘naturally’ belong
to each occurrence of the entity. Surrogate keys are identifiers that contain no inherent,
Logical Relational Data Modeling Standard Page 17 Property and Casualty Insurance Working Group
embedded data about the entity. That is to say, they are always non- intelligent keys.
Surrogate keys are usually a numeric attribute whose value can be generated
automatically either as a sequential number or a random number. Synonyms for a
surrogate key include: artificia l key, syn thetic key, arbi trary key, and sys tem- genera te d key.
A Foreig n Key is a primary key of one entity (the ‘parent’ or independent entity) that is duplicated in a separate, related entity (the ‘child’ or dependent entity). A foreign key is not
required to be unique within the child entity. A foreign key that is part of a composite
primary key in the child entity is known as an identifying or primary foreign key. Attributes
in a non- identifying foreign key become non- key attributes in the child entity.
Dimensional Data Modeling There are dimensional data modeling concepts such as the grain of the model, conformed dimensions, and diagramming layouts that deserve coverage in a standards document dedicated to dimensional modeling. The next few paragraphs talk about which parts of the Relational Data
Modeling Standard apply to the Dimensional Modeling Standards and which do not.
Relational Data Models are designed to support operational databases that capture complex information accurately. They deliver stable, flexible data structures that are easily navigated and can answer unanticipated questions. Dimensional Data Models are designed to support reporting and business analytics databases. They deliver simple, high- performance queries that answer a set of anticipated questions.
Although Relational and Dimensional Data Models serve different purposes, they share many of the same standards. Most importantly, they both use the Model Driven Architecture approach.
Also, the Logical Object, Entity, and Attribute Definition and Naming Guidelines apply to both styles of modeling. They are both Entity Relationship diagrams and both use the same IE modeling syntax. The Relationship Standards also apply to both though in practice relationship names are not used as often in Dimensional models as they are in Relational.
Logical Relational Data Modeling Standard Page 18 Property and Casualty Insurance Working Group
Dimensional models are a denormalized design. Super- types and sub- types would be merged.
Their diagramming layouts often use a star schema design and occasionally a snow- flaked design so their Diagramming Layout Guidelines are different from the Relational model.
Appendix Class Words The three tables below enumerate approved class words which come in three flavors: key words , u nits of measur e and objects . Each class word has a standard abbreviation, definition, and associated logical data type. The example is a typical column name and data value.
Key Word Abbreviation Definition Logical Example Datatype Amount AMT A quantity of money. NUMERIC Policy Face Amount = 1,200.0 Code CD Letters and numbers used STRING Sales Office Code for brevity to identify = AR11 something. Count CNT A numeric count or NUMERIC Active Employee calculated quantity of Count = 41,256 anything other than money, used when no unit of measure applies. Date DT Time stated in terms of DATE Disability Date = year, month and day. 2002/4/5 Description DSCR A statement that STRING Policy Change represents something in Reason Description words. = “Match coverage to changed income” Identifier, ID, ID Data that serves to STRING or Employee ID = Identification, uniquely identify one item 0123456 Identity in a group NUMERIC Indicator IND Data that can have only STRING Auditing Approval one of two values other Indicator = Y than NULL: Y(es) or N(o). (1 character) Line LN A set of characters STRING First Address Line normally printed or = “451 MAIN ST” displayed as one horizontal row. Name NM A word or words by which STRING Person Full Name = a thing is designated and “Sammy Somerset”
Logical Relational Data Modeling Standard Page 19 Property and Casualty Insurance Working Group
distinguished from others. Number NUM Normally numeric data STRING or Arrival Sequence used to identify ordinal Number = 5 position or to distinguish NUMERIC between items in a set. When numeric, it must always be a whole number. Objects See Object Binary Objects, such as STRING list below program objects, images, sounds, or videos. Percent PCT Numeric data specifying a NUMERIC Sales Closure Percentage portion or share out of Percentage = .7500 each 100 units. (75 units out of 100 is 75 percent (%). Percent values are multiplied by 0.01 in order to facilitate customary processing. In the example, 75 percent would be stored as 0.7500 but displayed as 75.00 %.) Text TXT Data having relatively STRING Audience Comment undefined content and Text = “Enthusiastic arrangement such as a and attentive” note, comments or an explanation Time TM Time stated in terms of TIME Check-In Time = hours, minutes and 8:45 AM seconds Timestamp TS Time stated in terms of TIMESTAMP Transaction year, month, day, hours, Timestamp = minutes, seconds and 20021203134516.8 fractions of seconds. 72 Identifies an instant in time. Units of See Unit of All units of measure, e.g. NUMERIC Measure Measure list Feet, Months, Miles, below Centimeters.
Logical Relational Data Modeling Standard Page 20 Property and Casualty Insurance Working Group
Unit of Measure Abbreviation Beats per Minute BPM Centimeters (Centimetres) CM Cubic Centimeters (Centimetres) CC Days DAY Degrees DEG Feet FT Grams G Horsepower HP Hours HR Inches IN Kilograms KG Kilometers (Kilometres) KM Kilometers (Kilometres) per Hour KMH Liters (Litres) L Meters (Metres) M Miles MILE Miles Per Hour MPH Millimeters (Millimetres) MM
Logical Relational Data Modeling Standard Page 21 Property and Casualty Insurance Working Group
Minutes MIN Months MO Ounces OZ Pounds LB Units. “Units” is a generic Unit of UNIT Measure (UOM) used when data with different UOM will be stored in a common column. In this case there must be a companion code column containing a UOM abbreviation indicating the UOM of the Units value. Weeks WK Years YR
Object Type Object Class Abbreviation C++ Program Object OBJ_C PowerBuilder Program Object OBJ_PB SmallTalk Program Object OBJ_ST Bitmap Image IMG_BMP Gif Image IMG_GIF Jpeg Image IMG_JPG Rav Sound SND_RAV
Logical Relational Data Modeling Standard Page 22