EB6546
EB6546
EB6546
4 Enhancing the FSDM with Unification During my tenure at this company, we purchased another
financial services company, and there was an extremely
5 Teradata FSDM Scenario
time-consuming, painful, and costly effort to integrate
8 Conclusion systems. Our company was eventually purchased by a
large global financial services company, and the integra-
8 About the Author tion issues were even more time-consuming, painful,
and costly. Having a solid enterprise view would have
definitely saved a substantial amount of time and money
Executive Summary during both of these integration processes.
7. Gracefully extends to accommodate new concepts. Rarely can traditional labels define organizations accu-
The FSDM contains a fair amount of abstraction. rately. That is particularly true in the current economy as
Abstraction means combining like things together many branch out to perform functions that cross tradi-
under generic terms, such as Event and Party, to tional industry and market boundaries. More and more,
facilitate integration, and to gracefully handle future companies simply do not fit any specific mold:
requirements. The FSDM can easily accommodate a
new type of Event, for example, as well as connect with Retailers may have a credit card line.
other Teradata iDMs that also use the Event concept. Car manufacturers may have a finance arm for loans.
This allows for greater commonality within and across
Oil and gas refiners may do oil and gas trading
the iDMs. All industries have Events, for example,
(commodity trading).
whether they are campaign solicitations in the bank-
ing industry, emails to change a shipment date in the Utilities may sell electricity (also commodity trading).
manufacturing industry, or service disruptions in the
communication industry. Abstraction has been applied Therefore, even when they know the benefits, diversified
in not just industry-neutral terms such as Event, but companies often say they cannot find a single suitable
also for terms specific for finance and insurance. For industry model that works best for them. Fortunately,
example, instead of specifying all of the dates associ- it is now possible to build a unified data model to see
ated with a claim, the generic concept of Claim Date beyond the enterprise view to the holistic value of an
Type allows for all types of dates to be easily accom- organization.
modated such as Occurred Date, Settled Date, Receipt
Date, and any new types of dates, as well. Companies typically choose an iDM that is tailored for
their industry. But to be at all useful, the model needs
8. Provides global and standard perspective.
to accommodate changes such as expansion into dif-
The structures on the iDMs are designed for international
ferent markets. That might require two or more iDMs
use, and not just United States-based. For example,
that must be either integrated or heavily customized. A
the term postal code is chosen over ZIP code. In
better method is to select one primary iDM and then add
addition, the data elements on each iDM follow best
the relevant components from other industries to cre-
practice naming standards, including the use of class
ate a unique model. For example, a utility company may
words based on the ISO 11179 metadata standard. A
choose the Teradata Utilities Data Model (UDM) as their
class word is the last part of a data element name that
primary iDM, yet leverage the modules of Account, Event,
represents the high-level category in which the data
Channel, Party, Geography, and Forecasting, to handle the
element belongs. Examples of class words are name,
commodity trading side of their business. This method
code, identifier, date, quantity, and amount. So the
of unification captures the best practices of industries
class word for Customer Last Name is name.
and creates a plug-and-play environment that employs
A medium-sized insurance company in the Midwest For example, in the automobile line of business, there is
primarily writes three lines of business (LOB): Auto, Resi- only a general classification used, which is an indicator
dential, and small commercial or Business Owners Policies capturing whether the claim was a result of human negli-
(BOP). Each business unit currently operates autono- gence or a result of Mother Nature. The business division
mously, and senior management has as a core business does have a more robust classification scheme than the
objective to integrate analytics and reporting for all three automobile LOB, yet it is done based upon character-
lines with the goals of more efficient operating processes, istics rather than events. So for example, if a hurricane
and more powerful and useful analytics information. causes damage to a small business that has its business
Today, even relatively simple reporting questions, such policy with this insurance company, instead of captur-
as How many of our auto insurance policyholders are ing the hurricane event, this insurance company captures
self-employed and therefore potential candidates for the characteristics of the incident such as the claim was
business liability insurance? can only be answered after a result of wind or rain. The residential unit does have the
much manual effort. Often there is a high degree of error in most robust way of capturing the cause of the claim, and
the results. Last year, for example, this insurance company they use Claim Reason Codes for this purpose. For exam-
conducted a direct mailing campaign to personal lines ple, a Claim Reason Code 55 is for a hurricane and a Claim
customers whom they believed were good prospective Reason Code 25 is for a broken water pipe. So right away
business liability insurance policyholders, offering them an even with this first example, there are major integration
attractive multi-policy discount. However, because of a issues. How can we answer the previously-raised question,
high degree of errors in answering cross-LOB questions, What is my exposure across all lines of business if there is
this insurance company mistakenly offered the multi-policy a Class 5 hurricane that will hit the southeast? if each line
discount to more than 200 existing business liability of business categorizes the cause of claims differently?
insurance policy holders who did not qualify for the offer.
Enterprise A communication tool for the Top 10 concepts and their business rules All at Once
Subject Area within the organization. Useful for explaining core concepts and for
Model explaining the similarities across the different business units.
Enterprise Although more detailed than the Enterprise Subject Area Model, this All at Once
Conceptual model still fits on a single piece of paper, and contains more industry-
Data Model specific terms, yet at a high enough level where the units fit together
holistically. Useful for communicating scope of large initiatives such as
the data warehouse program.
Integrated A very detailed and normalized view of the business, not tainted with Incrementally,
Data Layer technology such as software and hardware constraints. Useful for driven by project
Logical/Physical ensuring the data warehouse contains the information to answer busi- requirements
Data Model
ness questions, both questions within an LOB as well as cross-LOB
questions.
Integrated Data The integrated data layer PDM tainted with technology, such as soft- Incrementally,
Layer Physical ware and hardware constraints. Useful for working with database driven by project
Data Model administrators and developers to ensure the business questions are requirements
being answered with the most efficient structures and process. Along
with this model, we also need to have a spreadsheet containing the
mapping from source application data models to this model.
INTERNAL ORGANIZATION
A party that is a unit of business. CHANNEL
The vehicle by which a
LOCATION party may interact
A physical address, with the financial institution.
electronic address, CAMPAIGN
geographic or spatial A communication plan
area. to deliver a message.
complex logical design. The colors used on this model for Integrated Data Layer Logical Data Model
each subject area are also used for all of the entities within The integrated data layer logical data model for this
each subject area, for easy traceability. For example, all of example scenario was built incrementally on a project-by-
the Party entities in the FSDM are red in the Subject Area project basis. An in-depth business questions analysis was
Model, and also in the Conceptual and FSDM models. performed, and sets of business questions were bundled
into project deliverables. Many business analysts found it
It is typical to customize industry data models based on challenging to extract questions from the business folks.
need, and therefore even on this high-level data model Luckily, the FSDM and the Data Integration Roadmap came
some of the names and definitions need to be modified with several hundred typical business questions, which
for this particular insurance company. For example, this were used as a brain-storming technique with the busi-
insurance company prefers the term Sales Channel over ness to agree on a set of common questions. As you might
Channel, and so this entity was renamed on this model. expect, the integrated data layer LDM required more effort
than the prior two models. It had more detail and required
Enterprise Conceptual Data Model the most discussions to resolve the integration issues.
The conceptual data model contains more detail than the
subject area but yet still at a high enough level to be easily The FSDM was the starting point for each project, and the
readable on a single piece of paper. The FSDM also comes project teams often extended the model themselves with
with a conceptual data model, and this model contains the Architecture Center of Practice group reviewing any
about 200 entities, as opposed to the ten on the subject modeling work to ensure consistency.
area. Much of the additional detail on the conceptual
comes through subtyping, a small example shown for the
Event concept in Figure 3.
Teradata and the Teradata logo are registered trademarks of Teradata Corporation and/or its affiliates in the U.S. and worldwide. Data Model Scorecard is a trademark of
Steve Hoberman & Associates, LLC. Teradata and the Teradata logo are registered trademarks of Teradata Corporation and/or its affiliates in the U.S. and worldwide. ERwin
is a registered trademark of Computer Associates (CA). Teradata continually improves products as new technologies and components become available. Teradata, therefore,
reserves the right to change specifications without prior notice. All features, functions, and operations described herein may not be marketed in all parts of the world. Consult your
Teradata representative or Teradata.com for more information.
04.15EB6546