PERSONALIZED DATA SERVICE FOR MASTER DATA
MANAGEMENT
Jinchul Han, Min-goo Lee
Prompt, Inc.110 Yangjae-Dong, Seocho-Gu Seoul, Korea
Jonghoon Chun
Department of Computer Engineering Myongji University Yongin, Kyunggi-Do, Korea
Keywords: Master Data Management, Personalized Data Service, Composite Master.
Abstract: Recently, the needs for enterprise-wide Master Data Management(MDM) have been increased drastically.
MDM is the framework of processes and technologies aimed at integrating and managing the quality of
disparate enterprise master data to provide consistent information across diverse users and application
systems. Integrated master data environment may provide consistent, reliable, and accurate master data, yet
it may not be suitable for individual data needs. Utilizing view definition of commercial DBMSs is one way
to provide personalized data service, but it lacks flexibility in interoperability with other services provided
by MDM systems. In this paper, we propose a new model to make the data service personalized, flexible,
and efficient enough to use by physically combining underlying master data in order to meet individual data
requirements.
1 INTRODUCTION
As different departments and branches have
different data needs, they tend to build and use their
own proprietary systems. The systems and master
data used by them are suitable to their own needs
thus normally maintained separately. From the
enterprise-wide point of view, segregation of master
data and the application systems incur duplication of
master data, which will usually result in master data
inconsistency problem. Master data residing in
dispersed locations possibly with multiple copies
obviously makes the master data management task a
lot more complex, and almost impossible to
streamline and execute enterprise-wide business
processes and analyses. As a solution to it, MDM
has long been recognized as a means to guarantee
so-called “single version of truth” across an
enterprise.
MDM is the framework of processes and
technologies aimed at integrating and managing the
quality of disparate enterprise master data, which
almost always gets used by other users and systems
as one and only source of reference. The main
objective of MDM is to provide consistent
information across diverse users and application
systems by supporting sustainable framework for
data standardization environment.
Data integration with MDM can provide
consistent, reliable, and accurate master data to
multiple data consumers within as well as outside
enterprise. However, one may not be satisfied with
contents and formats of master data if only a single
static version of master data is enforced to be used
across the enterprise. One can imagine such cases
easily as people at sales departments may have
different master data views and needs from people at
R&D departments.
Such requirements may be fulfilled if one can
implement an MDM environment in which different
combination of master data are dynamically
composed and feed to the individual users and
systems as needed. We call that a “personalized data
service,” and believe that it is an indispensible
feature that an intelligent MDM solution must
provide in modern IT environment.
This paper presents a new model to support
efficient yet flexible management and reference
framework for personalized master data service. We
identify requirements for personalization and devise
technologies needed for it.
287
Han J., Lee M. and Chun J. (2009).
PERSONALIZED DATA SERVICE FOR MASTER DATA MANAGEMENT.
In Proceedings of the International Conference on e-Business, pages 287-293
DOI: 10.5220/0002227102870293
Copyright
c
SciTePress
This paper has four sections. Section 2 introduces
rudimentary concepts of master data and its
management. Also in section 2, requirements for
personalization are identified and contemplated.
Section 3 defines methods and technologies needed
for personalized data services. Section 4 concludes
with summary of the paper and presents further
studies.
2 MASTER DATA
MANAGEMENT
2.1 Master Data
Master data is information that acts as a key to the
operation of business and is the primary focus of the
Information Technology (IT) discipline of MDM.
This key business information may include data
about products, customers, financials, materials,
suppliers, employees, etc. which often are non-
transactional in nature. Master data is usually needed
by multiple functional groups and maintained in
various systems in dispersed locations within an
enterprise. Master data may not be maintained in a
single central repository; therefore, the possibility
exists for duplicated, inconsistent, and inaccurate
master data. Thus Master Data is that persistent,
non-transactional data that defines core business
entities for which there is an agreed view across the
organization.
Product Vendor
Product Vendor Material Su p p li er
Product Material Su p p li er
Syst em A
Syst em B
Sy st e m C
Figure 1: Master data used by multiple systems.
Figure 1 depicts a conventional system setup where
each system manipulates its own master data. As
seen in the figure, product, vendor, material, and
supplier masters are all redundantly maintained in
multiple systems, and there is no guarantee of these
copies of masters would agree on each other. This is
a very typical situation in modern IT system
environment. In some cases, system setups like this
take months to just generate a report on enterprise
wide spend analysis. Inconsistent, unreliable and bad
data results in bad quality report although it comes
in at a high cost, which in turn leads to a bad
decision making.
2.2 Data Integration
MDM is the framework of processes and
technologies aimed at integrating and managing the
quality of disparate enterprise master data, which
almost always gets used by other users and systems
as one and only source of reference. The main
objective of MDM is to provide consistent
information across diverse users and application
systems by supporting sustainable framework for
data standardization environment.
Data integration with MDM can provide consistent,
reliable, and accurate master data to multiple data
consumers within as well as outside enterprise.
Traditional master data integration processes are as
follows.
9 Analyze an individual system’s master data
9 Define an enterprise master data
9 Define an enterprise standardization of
classification hierarchy, attributes and UOM(unit
of measure)
9 Master data cleansing conforming to enterprise-
wide data standard
9 Master data integration
9 Continuous master data quality management
program running
9 Provide personalized master data services to
individual users and systems
VendorProduct Material Supplier
System A System B
System C
MDM
Figure 2: Data Integration.
Figure 2 shows an integrated master data
management framework where individual systems
can refer to. Essentially, MDM supports to share
consistent master data throughout the enterprise by
integrating and standardizing master data which,
before, were maintained independently by individual
systems.
Figure 3 is an architectural diagram of
eCliX
TM
MDM which is the commercial
implementation of our master data implementation
solution. eCliX
TM
MDM consists of three layers. The
interface layer is for master data interface to users
and systems, the business layer is for main
ICE-B 2009 - International Conference on E-business
288
functionalities of master data management including
management of master data and metadata, and their
quality management, and the storage layer is the
repository for storing and maintaining master data
and metadata.
Master data management module in the business
layer performs registration/modification/deletion of
master data. Metadata management module is
mainly for manipulation of metadata such as
classification hierarchy, attribute metadata, UOM
values, and information for audit trailing. Data
quality management module maintains quality of
master data by utilizing the information provided by
metadata. For example, if attribute metadata
enumerates standard values for a particular field of
master data, such information can be used to check
the validity of new entry.
Product
Vendor
Material
Supplier
System A System B
Personalization DIWorkspace
Metadata Mgt. DQ Mgt.
Master data
Mgt.
Rule Admin
General Library
Master Data
Metadata
Biz Layer
Storage
Layer
Interface
Layer
Figure 3: Master Data Management Architecture.
The first step toward enterprise-wide integrated
master data management environment is to figure
out master data from the entire set of data being used
by various systems of the enterprise. Master data
usually has distinguished characters that can be
differentiated from transactional data. Table 1 shows
a set of master data, namely product, vendor,
material, supplier masters and their attributes
maintained by individual systems. Example
attributes shown in the table are very typical ones in
real-life situations, as one needs careful
consideration whether to include which attributes as
master data attributes or not.
Generally master data are non-transient and
works as a single reference point for various
business processes. Any other data, that has
tendency to change a lot as time passes by, are
considered as transactional data. For example,
‘Name’ for vendor master does not change often,
whereas ‘Price’ would change a lot. Thus ‘Price’
would not be included as a part of master data
attributes whereas ‘Name’ would be. Obviously this
type of design decision is purely subjective and it is
up to the subject matter expert to provide sufficient
information to let the designer make rational
decision making.
From the implementation point of view, any data
that is included in the MDM repository are master
data, any others are transactional data.
Table 1: Redundant Master Data Attributes.
Product Vendor Material Supplier
System A
Name
Price
Date
Name
Rank
Price
Quantity
System B
Name
Date
Name
Name
Quantity
Name
System C Name
Name
Price
Quantity
Name
Rank
Separation of master from transactional data
from table 1 results in the following.
Table 2: Enterprise Master Data.
Master Data Transactional Data
Product Name, Date, Price
Vendor Name, Rank Price, Quantity
Material Name Price, Quantity
Supplier Name, Rank
Figure 4: Enterprise Master Data.
Table 2 and figure 4 together shows a set of
transactional data separated from master data.
Vendor.Price, Vendor.Quantity, Material.Price,
Material.Quantity are transactional data which needs
to be maintained independently from master data.
However, individual systems need these extra
transactional data to carry out their business
processes. Flexible integration capability is in need
not by system integration but by data integration.
We call that a “personalized data service.”
Vendo r
(Name, Rank)
Product
(Name, Dat e,
Pri ce)
Material
(Rank,)
Supplier
(Name, Rank)
Syst em A Sy st e m B
Syst em C
Vendor.Price
Vendor.Quantity
Material.Price
Material.Quantity
PERSONALIZED DATA SERVICE FOR MASTER DATA MANAGEMENT
289
In order to make personalized data service
possible, the following must be considered.
9 Ability to manage transactional data to later
combine with master data on the fly
9 Master data transformation based on
classification mapping, attributes, and UOM
9 Authentication, and authority control over users
and systems that uses master data
9 Adaptability to flexibly response to introduction
of new requirements and systems
In this paper, we propose a new model to resolve
aforementioned obstacles to provide flexible yet
efficient personalized data service environment.
3 PERSONALIZATION
Established MDM system can provide consistent,
reliable, and accurate master data to multiple data
users across enterprise. However, one may need
some fundamental transactional data as well as
master data in order to carry out his/her business
processes. For example, Vendor.Price, Vendor.
Quantity, Material.Price, Material.Quantity of figure
4 are examples of fundamental transactional data, in
this paper we call them additional properties of
master data. These data may be transactional over all,
yet they may well be very important master data to a
single system. Consequently, we need to handle
these transactional data differently from normal
transactional data and need to derive an efficient
method to integrate with enterprise-wide master data.
In this section, we propose a new framework to
efficiently provide information contents in the form
demanded by individual systems. In the framework,
physical view generation to integrate master data
with additional properties can easily be configured
by the user of the system.
3.1 Composite Master
Once the initial load of standardized and integrated
master data is complete, enterprise-wide master data
quality management program can launch. The new
program and system must not interfere with the
existing individual systems which were already in
operation. However, from the existing systems’
point of view, additional functional requirements
newly came up. One must be able to access the
MDM system in order to look up and use the master
data which before were locally maintained and used.
The view definition supported by commercial
DBMSs can be used to satisfy personalized data
requirements. However, users who require master
data, usually ask for additional information such as
classification hierarchy as well as code information
in a way that can be directly applied to their own
business processes. Clearly simple view definition
functionality of commercial relational DBMSs
cannot support this type of requirements. To do so
we propose new scheme based on the use of
composite master. Composite master is essentially a
personalized data view which are defined by
underlying unite masters. The users and systems can
dynamically create composite masters any time they
need for any purposes without having to directly use
bare SQL statement. By the system, any information
that is inherently associated with composite master
such as identification, attributes and classification
information are automatically reflected on composite
master.
Fig. 5 shows an example of the composite master
schema. System A needed a personalized data view,
which includes combination of information from
product and vendor as well as some additional
information. A composite master can be defined to
include name, data and price from product unit
maser, name and rank from vendor unit master, and
price and quantity as additional properties.
Although composite masters are physically
defined, actual values from unit masters are not
replicated. Only their reference addresses of
attributes and related information are physically
maintained by composite master definition. This
way, problems which may have been caused by data
replication, are naturally prevented from being
occurred. Additional attributes are defined and
added to the composite master in the same way as
other attributes are defined. Attributes are defined by
separate module -- property manager -- which allows
users of the MDM system to define new attributes
independently from any tables.
Vendor
(Name, Rank)
Pro du ct
(Name, Date,
Price)
Material
(Name)
Supplier
(Name, Rank)
Sy st em A
Pro du ct
Name Date Price
Vendor
Name Rank
Additional Prop.
Price Quantity
Figure 5: Example of Composite Master Schema.
3.2 Composite Manager
In order to establish personalized data service based
on the use of composite master, the following
requirements are identified.
ICE-B 2009 - International Conference on E-business
290
9 Ability to dynamically define composite master
9 Ability to define additional attributes for
composite master from existing set of attributes
as well as by defining new attributes on the fly
9 Flexible composition and integration with
underlying unit master data
9 New master data registration process must be
coordinated with the search/reference/validation
process of the existing master data
9 Attribute value and format transformation of
mater data in the form required by individual
systems and users
9 Independent from changes made to underlying
unit masters, but permanent deletion of master
data may need some extra work
Authentication and authority control for data
governance must be supported per individual unit
master
3.2.1 Composite Master Definition
Composite manager is a module responsible to
define and sustain composite masters
Product Vendor Supplier
System A
Composite Master
Composite
Master
Unit
Master
Composite
Manager
System B
Composite master
Material
System C
Composite master
Figure 6: Composite Manager Data Model.
Figure 6 outlines the functionality of composite
manager. Composite manager uses unit masters to
define new composite masters. First, composite
manager allows you to view and select predefined
unit masters. All or some of the attributes of the
selected unit masters may be used in definition of
composite master. Additionally if there are some
additional attributes needed while defining
composite masters, one can create new attributes and
use them as a part of composite master attributes. If
there are some attributes previously defined by using
property manager, one can select needed attributes
from the existing list.
The main function of property manager is the
creation of new attributes independently from any
relations. Some of the attributes may be defined
once and used by multiple unit and composite
masters. Some of the attributes may only be defined
and used as additional attributes for a single
composite master. Creation of new attributes
involves specification of constraints as well as some
of physical aspects of it. For brevity, in this paper,
we omit the rest of explanations of property manager.
Product Vendor Material
System A
Enterprise MD
Composite
Manager
Composite Master
Price, QuantityName, RankName, Price, Date
Supplier
Product Vendor Additional Property
System A
DB
Figure 7: Definition of Composite Master.
Figure 7 shows the schema of composite master. A
composite master consists of two parts – one for
storing reference information of underlying unit
masters, the other for storing additional attributes,
and it is physically saved and maintained. While
generating new composite masters, composite
manager works in coordination with authentication
and authority control module to block unauthorized
access to composite and unit masters.
3.2.2 New Data Registration/Modification/
Deletion of Composite Master
For efficient and consistent master data management,
standard procedure for new data registration to
composite master is needed. A person who has
access authority to composite master does not
necessarily have access authority to unit masters. In
other words, he/she only has the authority to
register/modify/delete additional attributes of
composite master, but he/she does not have the
authority to register/modify/delete associated
underlying unit masters.
The registration, modification, and deletion
procedure for composite master are as follows.
PERSONALIZED DATA SERVICE FOR MASTER DATA MANAGEMENT
291
Action
Enrollment
Modification
Deletion
Search/Select
Product
Select
Select
Process
Modify
Price, Quantity
Search/Select
Vendor
Delete
Write
Price, Quantity
Figure 8: Registration, Modification, Deletion Procedure.
Utilizing the two layers of masters, data stewards
can manage master data flexibly. A composite
master only references locations of unit masters
instead of maintaining copied values, it can sustain
independently from any changes made to underlying
unit masters. Efficient manipulation of unit master
reference addresses must accompany to ensure the
scalability.
3.2.3 Data Search in Composite Masters
In order for external users and systems to be able to
retrieve needed data from a composite master, unit
master search, transformation of the search results to
the representation that the user/system needs, and
dynamic composition with additional attributes are
needed.
Unit Master
Composite Manager
Composite Master
Metadata
Rule
System A
Search
1
Search
2
Search
3
Transform
4
5
View
Figure 9: Composite Master Data Retrieval.
Figure 9 shows the step by step processes needed to
retrieve master data from a composite master. The
following scenario breaks down the typical
procedures needed.
9 System A submits master data retrieval
requirements to composite manager (Step 1)
9 Search statement submitted to execute on
composite master(Step 2)
9 Decompose search statements into several
search statements of associated unit
masters(Step 3)
9 Search unit masters (Step 3 continued)
9 Search for rules applicable to the representation
format that system A requires (Step 4)
9 Transform the result according to the rules
found(Step 4)
9 Dynamically combine transformed results and
additional attributes(Step 5)
9 Send final result to system A(Step 5)
Figure 9 merely shows the simplest case retrieval
scenario. However, any information (e.g.,
classification hierarchy) that is related to the
composite master can be retrieved effectively and
efficiently in a similar manner. It is because such
information is automatically inherited from the unit
masters without having to have any additional
operations to redundantly specify.
3.3 Functional Requirements
In order to define and maintain composite masters,
the following functional and technological
requirements are needed. These functions and
technologies, in general, are not specific to our
approach, but also necessary for implementation of
master data management systems overall.
Table 3: Requirements Specification.
Technology Functionality Composite Mgr
CM Schema
Efficient definition
of composite master
Composite
Master
Definition
Admin DB
Access authority & authentication
definition and maintenance
Composite
Master
Definition
MD
Searching
Keyword & class search for
master data
Composite
Master
Register/Search
Rule DB
Storage & maintenance
of transformation rule
Master Data
Transformation
MD Monitor Monitoring of master data change
Deletion
Propagation
Table 3 summarizes main functions and
technologies needed for composite master
manipulation, and eCliX
TM
MDM is our commercial
product that implemented them.
4 CONCLUSIONS
In this paper, we investigated personalization issues
in the context of enterprise-wide MDM. As for
enablement of efficient personalized data service, we
ICE-B 2009 - International Conference on E-business
292
proposed composite master and composite manager
for management purposes. Composite masters are
different from conventional view definition, in that it
only maintains references to underlying physical
masters. Composite masters can also be defined with
additional attribues, and they provide extra
flexibility to individual users and systems. New
system deployment to MDM can no longer be any
problem, because users of new system can define
composite masters dynamically.
We believe that the concept of composite master
is very important for any organization that might
have some personalized data service requirements,
and soon it will be realized as an indispensible part
of master data definition.
Additionally, eCliX
TM
MDM provides
information retrieval capability to efficiently retrieve
any information that is related to given master data.
Quality management is another strong point of our
prouct. Enrichment of metadata makes master data
quality management process easy yet systematic. As
business metadata plays an important role in MDM,
enriched metadata repository and its management
functionality provided by eCliX
TM
MDM becomes
the necessity for any good MDM practice.
Composite master is still implemented as a part
of relational database schema physically realized.
Therefore, irresponsible deletion or modification
may complicate the constraints. Semi-automatic
propagation of composite master deletion/
modification needs to be further studied.
ACKNOWLEDGEMENTS
This research was supported in part by the Ministry
of Knowledge Economy, Korea, under the Industrial
Technology Research and Development Project
support program supervised by the Korea Institute of
Industrial Technology Evaluation and Planning.
REFERENCES
Hyun-gu, Jo. 2007. Achieving Value through Master Data
Management, Enterprise SOA Office, SAP Korea,
2007
IBM. 2008. Enterprise Master Data Management: Market
Review & Forecast for 2008-12. An MDM Institute
MarketPulse™ In-Depth Report
John, R. Andrew, W. 2008. The Seven Building Blocks of
MDM: A Framework for Success. Gartner Master
Data Management Summit.
John, R. 2008. MDM for Customer Data for Beginners.
Gartner Master Data management Summit.
Sung-won, Kim. 2006. MDM Strategy for Next
Generation Systems. SAMSUNG SDS Consulting
Review, No.4.
A, Z. 2007. Multi-Entity Master Data Management: The
4
Generation of MDM. An MDM Institute
MarketPulse™
Ted, F. 2008. Data Integration Technology and
Architecture: Increasing the Value of Your Master
Data. Gartner Master Data management Summit.
PERSONALIZED DATA SERVICE FOR MASTER DATA MANAGEMENT
293