Professional Documents
Culture Documents
Implementation
of
Data Model
ACORD CORPORATION
The ACORD Data Model, one of the seven facets of the ACORD Reference Architecture,
represents a comprehensive picture of all the major data entities that a carrier must
represent in its systems, regardless of domain (Property & Casualty, Life & Annuity, or
Reinsurance) or geographical scope.
The ACORD Data Model is a logical data model, not necessarily designed to be
implemented directly as a physical data model. In the course of using the ACORD Data
Model, data architects from insurance companies will have to make judgments on
whether and how to make certain denormalization decisions in order to best
accommodate their enterprise’s needs.
In addition, the ACORD Data Model has some data entities that are common across all
insurance lines of business (e.g., Agreement or Party), but also has some specific data
entities that may only be applicable to Life carriers or to Property & Casualty carriers.
Furthermore, the actual physical design of an insurance data model may differ depending
on how it is used. An architect may make different decisions if the data model is used for
a transactional system vs. an operational data store vs. a data warehouse.
This document is designed to provide a case study in which we will use the ACORD Data
Model to provide a suggested implementation for the staging of transactional data from
multiple disparate systems of a fictional insurer into an enterprise standard physical data
model.
The ACORD Data Model provides best practices developed with input from its member
and partner organizations. However, every organization must determine their own
approach to meet their needs and requirements. This document will provide guidance
and suggestions for assisting implementers in this process.
Hobbit Insurance is a 50-year-old midsize carrier serving primarily the Eastern US with a
subsidiary in Quebec and the Atlantic Provinces of Canada. It was originally founded as a
Personal Lines company, but now it also serves Property & Casualty needs for small to
midsize businesses – some “standard” products (BOP, Commercial Auto), as well as some
specialized lines of business (various E&O/Professional Liability lines, Kidnap & Ransom).
Because of several mergers in the past 20 years, Hobbit has at least two of every type of
insurance system (e.g., Policy Admin, Claims, Billing), and several smaller processes that
are maintained in complex spreadsheets or Access databases. The IT initiative that the
AS-IS Systems
Hobbit’s original and most used IT system is SAURON (System for Accumulating
Underwriting, Risks, Or Named insureds), which was originally developed for tracking
Insureds and their submissions and policies. It was enhanced over the years to provide
monolithic functionality across underwriting, claims, and billing for all the Personal Lines
parts of Hobbit’s business and some specialty Commercial lines.
The SAURON system supports personal automobile, homeowners, and umbrella insurance
through a single data model. These models establish a comprehensive data architecture
supporting the entire organization.
Hobbit Insurance continues to rely heavily on some systems developed by the companies
they acquired over the years.
One significant example is SMAUG (System Measuring Accounting & Underwriting
Growth). This system was acquired when Hobbit made its first major acquisition, when it
acquired a carrier focused on Commercial Lines, primarily covering small manufacturing
and small retail customers.
The third major source system that supports Hobbit’s business is SMEAGOL (System for
Medical malpractice, Errors & omissions, Architects and accountants, Gardeners, and
Officers Liability), the system that was acquired when they merged with a specialty
Professional Lines carrier. SMEAGOL provides functionality to support specialized lines of
Property & Casualty, primarily in the many subtypes of Professional Liability lines.
All three major systems cover the functionality to support the following aspects of the
Insurance business.
Customer Management
Sales and Marketing -- the submission/policy lifecycle
Policy/Coverage Administration
Claims Management
Financial Reporting
Needs Assessment for product offering for Personal and Commercial lines
o Home Insurance
o Renters Insurance
o Vehicles consisting of Automobile, Motorcycle, and Recreational Vehicles
o Windstorm Insurance
TO-BE System
All the Systems have some ability to generate management reporting for the lines of
business under their purview. However, the three systems have slightly different
definitions of the same concepts, and there is no non manual way to combine the results
across systems. This makes it very cumbersome when management wants a complete
view of the customer. Since the source of information resides in multiple different
databases, users extract data from each of the systems to consolidate reports in Excel for
management reporting.
Hobbit IT has been tasked with designing SAM (System for providing Answers to
Management). The objective for the new system is to create an integrated enterprise
wide database to ensure that there is a single version of the truth and a comprehensive
view of enterprise reporting. The data store will be the central repository for corporate
information representing the integrated data requirements of the enterprise and
designed to support the analytics, decision support, and reporting requirements of the
entire organization.
The data warehouse is characterized by:
Our approach is to design a data warehouse and structure it consistently with the business
area defined in the ACORD Information Model and Data Model. Many implementation
decisions need to be made before a final physical data model can be created. This
document will provide the approach used for the Hobbit Insurance use case. Depending
on your specific needs, your approach may vary, but most of the principles will be the
same.
The ACORD Data Model is a highly normalized view of insurance concepts. As the first
step, our approach was to establish an enterprise-wide conceptual data model for the
overall framework of Property and Casualty lines of business.
The following is the approach we used to building the data warehouse for Hobbit Insurance
Company. We started with a best-practice industry set of data models (the ACORD Data
Model). Then, we did the following steps:
Develop the Enterprise Model, taking advantage of the ACORD Data Model at a high
level. This required us to make decisions on how entities will be modelled, how to apply
subtype rules, how to resolve many-to-many relationships, and which additional
attributes are needed to support an enterprise-wide organization. In our approach, we
identified six major concepts and relationships of importance, which took into account
all seventeen subject areas represented in the ACORD Data Model. The ACORD Data
Model provided the needed 'blueprint' or picture of the target data structures that can
be used to establish a data architecture supporting the needs of an organization for its
current and future needs.
Develop the Business/Subject Area Models specific to the requirements for Hobbit’s
Property and Casualty business. Subject area models demonstrate how various options
were considered as making decisions on which aspects of the ACORD Data Model to
include. This also requires us to decide which aspects of the enterprise model are not
applicable to this application and remove those. For example, the Investment section in
the ACORD data model is primarily used for Life Insurance. Since Hobbit is a P&C
carrier, that subject area is of limited importance to us. The six subject areas where
Hobbit decided to focus its work were Party, Product, Policy, Claim, Finance, and
Reinsurance. Note that we combined multiple ACORD subject areas into one at times.
For example, both the Agreement and Physical Object areas of the ACORD data model
were collapsed into Hobbit’s Policy subject area.
Define Key Attributes, both logical primary keys and foreign keys. The ACORD data
model does not attempt to list all possible data elements for a given entity. Instead, it
focuses on just “important”, nearly universally common fields. For example, for the
Claim table, ACORD lists “important” data elements such as Date of Loss and Claim
Status, as well as a Primary Key of Claim Number and Foreign Keys such as Policy
Number or Claimant ID. The complete list of Claim data elements varies widely between
carriers, or even systems with a given carrier. Hobbit is using the important fields in the
ACORD Data Model to start its approach to coming up with a complete list.
This use case study is designed to provide suggestions to support implementers of the
ACORD Data Model. However, the guidance provided here should not be considered as
the only way to implement the model. Implementers should determine their approach
based on their unique implementation needs. This section covers some of the technical
considerations used while making these decisions.
Entity Definitions
Business concepts to tables: The ACORD Data Model allows you the flexibility to model
any situation. However, realistically, data modelers will need to model for specific
scenarios. This starts with determining supertype and subtype entities and considerations
for creating tables. Subject areas in ACORD Data Model consist of three main concepts.
You need to make decisions around how you would want to implement core entities and
then sub-types as well as code lists.
NOTE: Entities in the Data Model that have parent as “Information Model Object” are
the base concepts. These base entities are the foundation of conceptual model.
Table naming often corresponds to class naming in the ACORD Data Model. In some
cases, we did change names from the ACORD standard to some common names used in
the industry. For example, what’s commonly called “Policy” by P&C and Life insurers is
called “Agreement” in the ACORD Data Model. This was done by ACORD for two main
reasons: a) Policy would be misleading for a reinsurance company, and b) Agreement is a
supertype word that not only covers insurance contracts (e.g., policies or treaties), but
can also cover other types of agreements such as employment agreements or service
provider agreements. Since the focus of Hobbit’s project is just our insurance data, and
not enterprise-wide GL data, all our Agreements will be insurance policies, and thus the
Hobbit name is Policy.
Column names correspond to attribute naming.
Every table is provided a single column entity primary key id.
Figure 2.1: Exclusive Relationship: Party Name can be person name or organization name but not
both. This is resolved by having a name kind code
A non-identifying relationship is when the primary key attributes of the parent are
not part of the primary key attributes of the child.
Example: A book belongs to an owner, and an owner can own multiple books. But the
book can exist also without the owner, and ownership of it can change from one owner to
another. The relationship between a book and an owner is a non-identifying relationship.
(Stackflow, 2017, What's the difference between identifying and non-identifying relationships?)
A non-identifying relationship can be optional or mandatory, which means the foreign key
column allows NULL or disallows NULL, respectively.
Our first step was to use the ACORD Data Model to develop a high-level conceptual model
that identifies major entities and relationships of importance to the business for Hobbit
Insurance. The ACORD Data Model serves as a reference or core model that cover all
common data entities transcending all insurance business processes. The ACORD Data
Model represents:
The “top down” order in which we think about the subject areas is:
1. Party and all related logical entities
2. Product Specification for Types of Insurance - Auto, Home, etc.
3. Agreements covering all aspects of insurance policies for our Insureds
4. Claim and claim settlement
5. Finance entities to cover accounting Transactions and Payments & Receipts for
premium and claims settlements.
6. Reinsurance concepts from Hobbit’s point of view as a Cedent.
Our approach was to first identify master data business concepts. Much of the data that
needs to be the foundation of the data warehouse are the master data entities such as
Party (Insured and Broker), Products, and Locations (Contacts and Place). ACORD breaks
these into separate subject areas, but we at Hobbit will combine some of these with other
areas for simplicity as we have a focus on transactional data. There is also another section
in the ACORD Data Model called “Kind Code lists” which represent codes tables that we
talked about earlier. We will distribute these among the more business-related subject
areas for analysis purposes.
The enterprise conceptual data model was designed to represent the most important
data entities in major subject areas and their relationships with each other. This will
serve as the roadmap to present more detailed enterprise physical model as the next
step. The below picture (Figure 3.2) can be found in an accompanying pdf or in the
Erwin Model.
The property and casualty Data Model identify major entities and relationships of
importance to the business. The legend in the model diagram makes use of color to
highlight the major subject areas. Next section highlights the major features of each of
the subject areas in the Data Model.
Claim is tied to Party, Physical Object, Policy & Coverage and Reinsurance
through Agreement.
o Claim is tied to Policy through Policy coverage.
o Claim can represent compensation for covered loss based on policy
coverage, but need not be tied to the exchange of money as an FNOL
might not develop, or a reserve can be put up precautionary
The deliverables represent six subject area data models for our Property and Casualty
business to group all of our data entities into discrete areas to aid understanding.
Well, let me think. Since I’m a P&C primary insurer, there are a bunch of ACORD organization
roles that are outside of my scope – e.g., Beneficiary, Guarantor, Annuitant, etc. Furthermore,
while I can imagine that my systems deal with many possible roles, for most of them, I really
only care about a limited set of information like name, and address, and phone/email. So which
types do I really care about having detailed data elements? Probably Insured, Broker, Reinsurer.
So, I will have a Party table, with a bunch of children tables, such as PartyAddress, PartyEmail,
and PartyPhone to allow for multiples per Party. For my developers’ sake, I will denormalize the
primary address and phone and email onto the Party table. I’ll create separate tables for
Insured and Broker and Reinsurer to house specific data elements. These will have the PartyID
as an FK on them. I’ll also create a PartyType table which is many to one with Party to allow for
common cases where a Party serves in multiple roles (e.g. both an Insured and a Claimant)
As discussed earlier, there are dozens of subtypes of Party that may be used in our
Hobbit Insurance data model. There are only a handful of subtypes where we may store
information other than name and address basics.
INSURED is probably the most important, and we will model that as a separate table as
Insured represents the customer view. In addition, there are enough different details
that need to be captured for types of individuals (INSURED PERSON) and organizations
(INSURED ORGANIZATION) to warrant the creation of separate tables. Also, Insured
contains other related reference entities such as INSURED DETAIL to capture financial as
well as non-financial relationships such as Accounts, Affiliations, Group, preferences,
etc.
PARTY is a catchall subtype that will be used for all other subtypes (e.g., MGA, TPA, Law
Firm, Witness, Police Officer) to capture basic name information. It will also have an
entry for the five specific subtypes mentioned in the previous two paragraphs. Other
many-to-one children entities such as PARTY NAME, PARTY ADDRESS, PARTY PHONE,
PARTY EMAIL ADDRESS, will be created for use for both the five specific subtypes and all
the generic subtypes.
We have used the following code lists in our model to demonstrate usage of reference
data. In our examples, we are using ACORD provided code lists. As an organization, you
may want to evaluate your master data strategy to include code lists, taxonomies, and
hierarchies of data.
Below are relationship of CORE Entities for Party and Contact & Place
The ACORD Data Model provides a supertype “product base” and many subtypes that deal
with various types of coverages, provisions, limits, etc. While this may provide a more
comprehensive design for some organizations. I really only care about the most important
ones such as Coverage, Deductibles and Limits. My model will have a Product Specification
table to store core attributes that define the product and it would have master data
relationships to define Broad Line of Business, Lines of Business as well as other code lists. This
would provide core attributes to be used in analytics. I have decided to create child tables that
deal with core components to identify coverage, limit and deductibles. I also decided to create
a separate entity to capture “All other” types of components in a single table having type
codes -- this will give me the flexibility to model things I cannot predict right now without
dramatically changing the model.
Product Specification is reference data which insurance companies can use to manage
their coverage and rules used for constructing policies. The ACORD Product specification
model provides extensive functionality to specify products and rules that can be applied
to technical design such as attribute rules, or rules that can be part of coverage such as
calculation rules. For this phase of the design, we decided to focus on core entities of
product specification and did not implement rule specifications.
ACORD provides structure around three main areas when providing elements to
construct product offerings
We have used the following code lists in our model to demonstrate usage of reference
data.
Property and casualty insurance are specifically designed to help protect your possessions
from theft or destruction. It also may protect your assets from being depleted through
disaster or litigation claims brought against you. Many organizations set up Product
Specifications as master data to describe coverage types, amounts, and provisions used in
selling policies to insureds.
Policy structure is aligned with the product structure and definitions. In addition to the
entities referenced in product specifications, the Policy subject area also covers the
ACORD Data Model section of Physical Objects. Property policies provide coverage for
physical items, such as homes, commercial buildings, motor vehicles, and personal
possessions or business inventory. Types of property insurance include homeowner’s
insurance, fire insurance, flood or earthquake insurance, and automobile insurance.
These insurance policies may have an “open perils” or a “named perils” clause. The open
perils clause covers losses for reasons that are NOT specifically excluded in the policy,
such as earthquakes, floods, and acts of terrorism or war. A named perils clause requires
the actual cause of loss (Peril) to be listed in the policy, such as windstorm, tornado,
pandemics, flood, and theft. For example, depending on the named perils included, water
damage may be covered from a water heater leak, but if the same water damage were
caused by a flood from hurricane, it may not be covered.
Casualty/Liability coverage covers you for losses that you may cause to another individual
or business. For example, if you have liability insurance on your car and another party is
injured in a collision caused by you, your liability insurance will take care of the other
person’s medical and repair costs. In addition, if someone sues you because of harm you
may have caused to him or to his possessions, your casualty insurance may cover the cost.
Both individuals and businesses can purchase property and casualty insurance. Personal
policies include homeowner’s insurance, renter’s insurance, and automobile insurance,
whereas commercial policies are written specifically for businesses and other
organizations and may include commercial general liability, and commercial property
insurance.
Narrative
We are making similar decisions in the Policy area around Coverages, Limits, Deductibles, and Clauses
that we made in the Product Specification subject area. We view the Policy as an instance of the
Product Class so they will have parallel structures. In the model, we are aligning the Policy structure
the same way. First, we created a Policy table that would hold core single valued attributes that are
relevant of all lines of business (such as effective date, policy number, date bound) as well as well as all
foreign keys and associations (such as underwriter, insured, broker, various lines or classes of
business). We also decided to instantiate core components as separate tables such as coverage, limits,
deductibles and clauses. Since we are building a data warehouse, we don’t need to create separate
We are modeling agreements for Insurance Policies and Reinsurance Agreements separately. The
ACORD Data Model treats them both as Agreements, which would be a good design if Hobbit were a
combination insurer and reinsurer (as many Bermuda companies are). However, at Hobbit, we only
sell insurance and only buy reinsurance, and we always do gross, ceded, and net calculations. Thus,
we will treat Insurance Policies and Reinsurance Agreements as two entirely different physical tables in
two different subject areas, even though ACORD considers them subtypes of Agreement.
Another area of focus was how we chose to model Physical Objects we are insuring. We decided not to
create just one physical object supertype table that would store all types of information for buildings
as well as vehicles, watercraft, livestock, jewelry, etc. This table, if instantiated, will be one
humongous table having lots of columns for all types of objects in the coverage and you would
constantly be adding more columns as needed. We evaluated our business and realized our
underwriting activity weighed heavily towards automobiles and water boats as well as houses,
businesses, structures, etc. Our focus on object subtypes was to instantiate entities for insurable
objects that require specialized information. Our decision was to instantiate building, land vehicles
and watercraft to store attributes that are very different from each type of object. For building we
may information about type of roof, how many floors, etc. For land vehicles, we can have how many
passengers, make, model and year of car. Of course, watercraft attributes are even different from
land vehicles. We can also make use of a generic Physical Object table if we need to keep track of just
basic name and description information (e.g., jewelry or fine art).
NOTE: In the diagrams, we show a specific coverages for leisure aircraft, we did not actually
instantiate this table, but we did include this in the subject area model for illustration purposes. If
there is a need to create specific physical object entities in the future, this is how it would look like.
For example, if we start insuring thoroughbred horses, we could make a separate Horse table with
specific data elements such as lineage, trainer, jockey, and medical conditions, and divorce it from
the generic Physical Object table.
Our supertype Physical Object has unique key ids that are referenced on all the subtype type tables.
I’ll create separate tables for Land Vehicle and Watercraft and Building to house specific data
elements. These will have the PhysicalObjectKeyID as an FK on them. One of the reasons we took this
approach was to support data warehouse reporting functionality. Having Physical Object table will
help having customer focused information where developers can easily aggregate information by
policy or by physical object characteristic.
Policy is the focal point for all insurance contracts. It is composed of Policy components
and rules. Tables in Product Specification and Policy share the same structure because
they store the same information. The difference between the two is that in product
specification, we store all possible offers that can be presented to a client, whereas in
Policy, it is the exact policy assigned to a client.
Example:
Bilbo has coverage for Store A, Warehouse B, and Office C at International Drive in Bag
End which covers fire and water damage. However, if the fire is caused by an
earthquake and the peril is clearly stated in Policy Peril for exclusion, there is no
coverage. In addition, he will have coverage for personal property for Store A, where he
is operating his own business, which covers any machinery, desks and office equipment
in the building. He is renting Warehouse B, so, in addition, he has a Policy Clause to
exclude property coverage for Warehouse B because the renter is responsible for buying
his own insurance for his personal property. He can also have a clause for coinsurance if
loss was due to renter’s negligence in Warehouse B. Subtype components can be
applied to Policy or Policy Coverage level of granularity. Policy Deductible, Policy Limit,
and Policy Clause have an FK to the Policy table that is nullable to allow relationship to
either Policy or Policy Coverage. For example, Limit Component of Policy Coverage,
Bilbo can have LIMIT on Coverage of $100,000 on Property in Store A. He can also have
another maximum limit on the Policy, in case loss is to both buildings, his business and
renter’s business of 1M, based on the assessment (Policy Assessment) of the building.
His policy can also have a Policy Deductible amount that is uniform everywhere.
Results:
Insured (1 row) -- Bilbo Baggins
Policy (1 row) -- BOP policy
Policy Coverage (3 rows) -- Property, Commercial Auto, Liability
Policy Limit (6+ rows) -- $1M overall limit, $500K per claim limit (attached to null
coverage), $800K property limit, $500K CA limit, $500K liability limit, $100K limit on
property in Store A
Policy Deductible (1 row) -- $1000 per claim deductible
We have used the following code lists in our model to demonstrate usage of reference
data. We recommend usage of code lists and master reference data is one of the master
data subjects every organization needs to decide how to implement.
Narrative
The claim area covers the major entities defined in claims management. The process
begins with an event resulting in one claim or multiple claims. Claim offers represent
multiple versions of claim details as they are being investigated and negotiated. Claim
settlement represents the financial movements as reserves and payments are being
decided. The relationship of claim coverage to policy coverage is shown as these are the
losses covered under the claim are attached to the appropriate sections of the policy.
The ACORD Data Model has a relationship between Loss Event and Loss Occurrence. A
Loss Event may be unique to a particular claim (e.g., a fender-bender) or may be
recognized by the industry as a whole (e.g., a named windstorm). The concept of a loss
occurrence is the basis of one or several claims (e.g., multiple claims may be filed for a
single loss occurrence, as when a claim is filed by each driver in a fender-bender). The
loss occurrence is characterized by a date, a location, a type, a potential cause, and the
eventual circumstances. For our model, we decided to implement a Loss Event table
which is optional (Loss Event Key is nullable FK on Claim Table), the purpose is to capture
event and occurrence information in a single entity.
NOTE: In the subject area model, we do show a relationship between Loss Event to Loss
Occurrence and Loss Occurrence to Claim for reference. This may be important for
operational system however, for Hobbit’s purposes we did not need this level of
granularity.
The Claim subject area will cover topics that associate parties with claims, policies with
claims, and claim reserves and payments.
Claim is tied to party, physical object, and policy coverage through foreign keys. A claim
pays for benefits or losses covered in a policy. Claims are validated, then the carrier may
present different offers for the claim, and once approved, the carrier makes payment to
the insured or the impacted party.
Eggshell Eddie was driving his Volvo with his two children, Robin and Wren. As he
proceeds through an intersection on a green light, defendant Crazy Craig runs a red
light and strikes Eggshell’s vehicle causing serious injuries to Eggshell, Robin and Wren.
After striking Eggshell’s car, Crazy’s vehicle impacts vehicle number two that was
stopped at the intersection waiting to turn left. Vehicle two was operated by Professor
Timmy Tyler. Eggshell Eddie, Robin and Wren all suffered serious injuries. Professor
Tyler suffered relatively minor injuries.
Crazy Craig was insured by Hobbit Insurance Company under a $50,000 per
person/$100,000 per accident automobile insurance policy. Crazy had no assets.
(Based on Copley, R. K. (2012). Insurer’s duties when policy limits demands involve multiple claimants.
Para. 1)
Let’s view this claim to see how it relates to the model for Hobbit and how we decided to
implement it using the ACORD Data Model as the basis.
Claim 1: Eggshell Eddie is the claimant and files a claim with Crazy Craig’s Insurance
Company (Hobbit Insurance Company).
Below are some of the key decisions on how we modeled the Claim subject area for
Hobbit.
Loss Event: A loss event is the base of one or several claims. We decided that we are only
going to use loss event for industry recognized events, such as named catastrophes like
hurricanes or earthquakes or pollution events. As this is not the case here, the nullable
Loss Event id on both Loss Occurrence and Claim can remain null.
Loss Occurrence: A loss occurrence is the base of one or several closely related claims.
We decided that it is not necessary to physically instantiate this table. In our scenario, we
could have three claims being generated from the one accident, but we will have an entry
in the Loss Occurrence table. However, it’s unlikely that our source systems tie these
three claims together as they could have come on different times (in in practice, one or
more of the claims may be going through another insurer). Thus, the difficulty of tying
these claims together and the limited value that can be gotten from that leads us to the
decision that we don’t need the table.
Claim: Claim is the base table to store details of claim. Policy ID is associated to Claim as
an FK in Claim.
Party Roles in Claim: Various party roles such as claimant, driver, and witness will be
associated to Claim. We decided to have Claimant as a table because there is specific
information related to this role. If there is a need to identify other parties (witness)
involved in the claim, Other Party Roles table will be used along with Party Kind Code List.
Decision on Party Role in Claim
o Witness
o Claimant
o Claim Adjuster
o Driver
o Etc.
We have used the following code lists in our model to demonstrate usage of reference
data. We recommend usage of code lists and master reference data as one of the master
data subjects every organization needs to decide how to implement.
The claims section below defines the major entities involved in claims management,
beginning with an event resulting in claims. As with other major business areas, the
relationships and party roles for claims is defined for the major entities. The relationship
of coverage to claim is shown along with financial aspect related to claim settlement.
Relationship Relationship
Parent Entity Name Child Entity Name
Name Type Relationship Definition
This relationship links a claim with
Claim ON Claim a corresponding claim coverage
Coverage HB_Claim HB_Claim Coverage Identifying component.
This relationship links one or more
claim offers to a claim to
Claim ON Claim represent the composition of a
Offer HB_Claim HB_Claim Offer Identifying claim offer.
Claim ON Claim This relationship links a claim to a
Status HB_Claim HB_Claim Status Identifying status at a point in time
Claim ON
Claimant Party Non- This relationship links a party role
Role HB_Claim HB_Claimant Identifying (Claimant) in a claim
Claim ON Other Non- This relationship links a party role
Party Roles HB_Claim HB_Other Party Roles Identifying (Other Party Roles) in a claim
Claim Coverage
ON Claim HB_Claim Coverage This relationship links a claim
Coverage Status HB_Claim Coverage Status Identifying coverage to a status.
Claim Coverage This relationship links the result of
to Claim an assessment and the physical
Assessment HB_Claim_Assessment Non- object that served as the subject
Result HB_Claim Coverage Result Identifying of the assessment.
This relationship links a claim
provision with a claim coverage.
This relationship determines how
Claim Coverage limits and deductibles are utilized
TO Claim and determine the actual claim
Provision HB_Claim Coverage HB_Claim Provision Identifying amounts
Claim Offer ON
Claim Offer This relationship links a claim
Status HB_Claim Offer HB_Claim Offer Status Identifying offer to a status.
This relationship links a claim
component to a related claim
ClaimComponent component. For example, a claim
Comprises HB_Claim HB_Claim Non- deductible may be associated
ClaimComponent Other_Components Other_Components Identifying with a claim coverage.
This relationship links a claim
Claim Provision provision with a financial
TO Financial Non- provision that actually hits the
Provision HB_Claim Provision HB_Financial Provision Identifying ledgers of Hobbit.
This relationship links a claim
Reinsurance provision with a Reinsurance
Claim Provision Claim provision, This is when it is
ON Claim HB_Reinsurance Claim determined that claim is linked to
Provision HB_Claim Provision Provision Identifying reinsurance agreement
Claim
Assessment HB_Claim_Assessment HB_Claim_Assessment This relationship links an
Result ON Claim Result Result Status Identifying assessment result to a status.
Finance is a common subject area that represents the financial amounts in context of Policy
amounts (premiums, fees, etc.) or claim amounts (claim settlement/payout). This subject area
covers two subareas – financial transactions that are assets and liabilities (e.g., premiums,
commissions, taxes, fees, surcharges, claim reserves, claims paid losses, IBNR, etc.), and the cash
flow transactions (e.g., premium coming in, losses going out, invoices, write-offs, etc.).
Narrative
In an organization, the Finance function covers many aspects such as paying employees,
receiving money from revenue streams, accounting and general ledger management,
Operating expenses, Corporate Balance Sheets. For the SAM project at Hobbit, I am
interested in only considering insurance related transactions. Thus, I am interested in
underwriting revenues that come from cash collected on policy premiums and money paid out
on claims. This SAM system will not cover non-insurance related money streams – payroll,
rent, investment income, for example. To get a full picture of Hobbit’s financials, Dr. Took will
have to look at reports coming out of the General Ledger system, which may be an extension
to this project someday. Also as noted before, this system is focused on reporting results as
they were entered – it is out of scope to be able to re-create the exact logic for why a
premium was $1000 rather than $1200, or why a large claim payout was $93,000, not
$104,000. As a best practice, we do not want to recalculate any financial information coming
from the operational systems.
The Insurance Revenue streams for Hobbit Insurance are the revenues generated from
premiums paid by insureds, plus claims paid to Hobbit from Reinsurers. Monies paid out from
Hobbit would be settlement for claims and premium share for reinsurers. Essentially, these
are just the monies coming in and monies going out. (Note this includes Reserves, which have
not actually gone out yet, but are earmarked for going out). We created a Financial Provision
table for storing details about all transactions resulting from these four activities. This table
will have premium, commissions, taxes, fees, surcharges, reserves, subrogation, claim
payments amount, reinsurance payables, reinsurance receivables, etc. The ACORD Data
Model provides an option to create separate entities for Financial Provision such as Fees,
commissions, premiums, taxes, etc. These entities might be needed in an operational system,
due to the complexity of calculations and rules that need to apply at that level of granularity.
As mentioned above, our objective is to store the relevant results (we don’t need to show our
work, just the answers). We can use this supertype table that can differentiate the types of
entry using the code lists.
Even though Financial Provision is the core table, we needed other tables to store details that
are geared towards our reporting needs. The ACORD Data Model has a supertype table that
As opposed to asset and liability transactions above, we decided to split off the subtype tables
of Financial Transaction for all Cash Flow activities. We decided to store Payment and
Payment Due, as these are the actual payment transactions (Money Received and Money
Sent) into a single table called Financial Transaction. We did instantiate a separate table for
Payment Settlement used for reconciliation between Financial Provision and Financial
Transaction.
Finance is a subject area consisting of financial accounting (premiums and losses) and
money transactions such as premium receipts or claim payments. The focus for the
Hobbit Data Model is to address Finance in context of Insurance related transactions
only. We will not be incorporating non-insurance General Ledger or Investment aspects
into this model.
Claim Provision: Claim provision is the place where we store ground up losses and
reserves for claim settlement. Claim Provision will store information about claim
payment as well as components such as deductible eligibility or limits applied as per the
Policy. For example, an auto policy might have a $1000 deductible and clauses around
what types of losses are covered. When the First Notice of Loss comes in, we may not
know the exact amounts of the loss or whether the investigation will show that the
coverage conditions are fully met. We may put in a placeholder row in Claim Provision
saying it is a $5000 rough loss. Later on, if we determine that half the $5000 was not
covered and that the deductible did apply, that $5000 Claim Provision may translate to a
$1500 loss in Financial Provision which is the actual amount that we pay. There may be
cases where there is no Financial Provision row because the claim amount did not
exceed the deductible or the loss was not covered at all. Claim Provision is a child entity
to Claim Coverage since actual compensation for the claim is determined by the
coverage.
Looking at the ACORD Data Model with the large variety of subtypes on Financial
Provisions, while we decided to split the insurance and reinsurance subtypes into
different physical tables, we decided not to split the individual transactions types (e.g.,
Premium, Commissions, Paid Losses) into separate tables as nearly all reporting needs
mix those subtypes. Instead, we will have a Provision Type code list on the Financial
Provision Table. We also decided not to implement the ACORD Financial Provision
Summary, as information can be summarized using Financial Provision details as seen in
the example below.
Here’s the relationship between Financial Provision, Payment Transaction, and Payment
Settlement.
Business Event 1 – Let’s suppose that Hobbit sells a BOP Policy to Nick’s Bagel Shoppe
that has coverage for Property and Commercial Auto, totaling $10,000, split evenly
between the two coverages. Of that $10,000, $1000 goes to Sean the Broker for
commission, $500 goes to the State of New York for a combination of taxes and fees,
and the remaining $8500 is our written premium. That would get entered as six rows in
Financial Provision (three provision types for each of two coverages, it would be more if
the policy were set up with installments).
Business Event 2 – Hobbit would then generate an invoice to Nick for the entire
$10,000. This would be a single row in Payment Transaction with a type code of Invoice
(the invoice lines behind that invoice would be the Financial Provision rows.
Business Event 3 -- Nick would then send us an EFT for $10,000. This would be a single
row in Payment Transaction with a type code of Payment.
Business Event 4 – Somebody at Hobbit would have the job function of matching the
payment to the invoice. That would create six associative rows in the Payment
Settlement table, each of which would be marked as fully paid.
Business Event 5 – In another business event three months later, elderly customer
Howard carelessly breaks the glass display container in the Shoppe by repeatedly
banging on it with his cane. Nick files an FNOL stating that he thinks it will cost $8,000
to fix. Hobbit enters a single row in Claim Provision (one coverage, one claimant, one
date, a reserve)
Business Event 7 – Hobbit mails Nick a check for $11,000. This adds a negative reserve
in Financial Provision, adds a Paid Loss row in Financial Provision, creates a row in
Payment Transaction of type code “Payment”, and creates a row in Payment Settlement
linking the Payment to the Paid Loss.
We have used the following code lists in our model to demonstrate usage of reference data. We
recommend usage of code lists and master reference data as one of the master data subjects
every organization needs to decide how to implement.
HB_Claim_Provision Claim provision is the financial aspect of money composition for claim
settlement. For example, Claim Provision will store information about claim
payment as well as components such as deductible eligibility or Limits applied
as per the Policy. Based on all the elements of the composition, Claim
provision may result in a Financial Provision where money will be exchanged
to settle the claim. There may be a case where there is no Financial Provision
because the claim amount did not exceed the deductible or limits rules.
HB_Payment_Schedular While Financial Provision represent the definition of the amounts and their
composition, the money scheduler defines how the money will be paid, when
it will be paid and who will pay and receive it.
HB_Payment_Transaction A transfer of money between two parties. This represents payments made or
payments received.
HB_Payment_Invoice_Header These details are derived from the money provision element parts that are
attached to the money provisions for this money scheduler.
HB_Payment_Settlement Payment settlement represents the cash application to match payment due
from Financial Provision and Payments received from Financial Transaction.
The Reinsurance subject represents agreements between a Cedent and a reinsurer. The
two basic methods of reinsurance are Facultative and Treaty. Within the scope of each
of these two methods, there are two broad types - Proportional reinsurance and Non-
Proportional reinsurance.
Facultative reinsurance is a method, under which the ceding company or Cedent offers
each risk individually to the reinsurer for its consideration. The reinsurer receives full
details of each underwriting policy offered, and has the option to reject the offer or to
accept.
Treaty reinsurance is an obligation based on an agreement concluded for a specified
period covering a defined set of risks (e.g., all homes located on Cape Cod against
windstorm risks). The Cedent is obliged to cede all risks falling within the scope of the
agreement, and the reinsurer is obliged to accept all such cessions without the option to
decline or to ask for improved terms. Offer and acceptance are thus automatic.
Definition of proportional reinsurance, the Cedent and the reinsurer share the risk and
they also share the premiums and claims in a set proportion. This concept could be
applied to an individual risk (facultative) or to several risks covered by a single contract
(treaty).
Narrative
As a P&C insurer, Hobbit’s product line deals with commercial and personal property
insurance coverages for personal and commercial business. Our insurance covers losses and
damage to residence and commercial lines in the eastern part of the US and Canada, as well
as several other states elsewhere. The risk of losses due to catastrophic windstorms is much
higher on the coastline than places like Ohio. Some of the high-risk policies that are written
on the Eastern Seaboard are reinsured with other carriers.
The ACORD Data Model provides a way to model underwriting of policies, facultative and
treaty agreements. For the data warehouse project, my focus is to capture accurate data on
both inward policies and outward treaties and fac certs, as well as to keep track of all
financial transactions related to any inward or outward contracts, so that I can do Gross,
Ceded, and Net reporting at any degree. I don’t need to capture all the elaborate rules for
what criteria makes a given Hobbit policy attach or not attach to a given treaty – I will
assume that the master data about the contracts and the transactional data about the
financial results is the core to my needs.
The concept of insurance risk means policies can exist without reinsurance and we, at
Hobbit would bear the full 100% of obligation. However, because our risks can be
correlated (for example, a Hurricane Sandy type of event would have wiped out decades
of underwriting profit if we had had to bear it without benefit of reinsurance),
reinsurance means that our risks are distributed across a wider basis. Reinsurance starts
with a simple concept where the reinsurer gets a certain amount of premium from its
Cedent (insurer) and shares in the Cedent’s risk in an agreed upon proportion or on an
agreed amount. In our Data warehouse, we are not concerned with the process of how
the reinsurance contract is created and managed on a day to day basis and how you
decide which policies are covered under an agreement. Our focus was to capture
details of the agreement that can be reported on and to know which policies and claims
are subject to reinsurance arrangement.
Bag End Brokers is a reinsurance broker to a ceding company (Hobbit Insurance). As the
Cedent’s broker, Bag End secured a reinsurance group composed of several property
and casualty reinsurers, including Rivendell Re, which has agreed to serve as the lead
reinsurer in the consortium. Other respected members of the reinsurance community
such as Lothlorien Associate, Gondor Re, and Rohan Re. The members of the
participating reinsurance companies entered into a reinsurance agreement through
which the reinsurer participants agreed to reinsure the risks covered by certain
insurance policies issued by the Cedent (insurer- Hobbit). Terms of the contract are in
the reinsurer agreement.
Decision on Reinsurance Agreement
Our focus for the reinsurance subject area was to capture the reinsurance agreement
and participant details. From the ACORD Data Model Agreement supertype, we created
a Reinsurance Agreement table to store details about the overall reinsurance contract
terms. contract with reinsuring property and casualty insurance companies that are the
reinsurers. For share of participation, Reinsurance Share was created to store either
percentage or amount of risk of all the participants. Reinsurance Share is a child table
with FKs to Reinsurance Agreement and Reinsurer (Party) for all the participants in the
agreement. In a reinsurance arrangement, there is a many-to-many relationship
between inward policies and reinsurance agreements. To accommodate this, we
created a Reinsurance Policy Mapping table that will list all the inward policies with
what outward agreements cover them, as well as all the individual policies that are in
the purview of a given outward agreement. Creating this mapping table facilitates the
below information in the Data warehouse
DDL script
A case study of applying this approach for modeling Property and Casualty line of business for
an insurance system was given. A practical strategy for transforming ACORD’s conceptual highly
normalized model into a business specific physical model was presented for the use case.
We also indicated the most important task is to validate the interpretation and transformation
of the ACORD model. This effort is lengthy; a designer needs to evaluate and make decisions
entity by entity and relationship by relationship using real-life business scenarios for their
organization.
However, each organization must determine their approach to meet their needs and
requirements. This document is designed to provide guidance and suggestions for assisting
implementers in this process.
We have presented the details of the data model and following supporting information as
separate documents
Erwin Data Model elements
DDL scripts for Hobbit Data Model
PDF of Data Model Diagrams
PDF of ACORD Code List Values (
NOTE: Code List Names are provided in the erwin Model. However,
this report provides the enumeration values in the ACORD Data
Model.