1. Brotherhood of Master Data - The MDM Cult Series Part 1
Dears,
The reason why the title has the world “cult” is to know that MDM is still not fully explored
technology and successful execution of the MDM projects remains the authority of little
knowledgeable brotherhood who indeed keep this secret very close to the professional network.
This is an attempt to unleash the knowledge of MDM to wide group of audience whom I feel would
make the best use of this information in their MDM Initiatives.
This article is designed to give a business-centric view of the master data management (MDM)
concept and what challenges will be faced by organizations of various sizes. Business stakeholders
within an organization will gain an understanding of the needs of their organization and will be
empowered to define the initial MDM vision for their organization. This vision will facilitate the
initial decision necessary to implement a master data management solution within their
organization.
Different Challenges based on size of Organization
As any business grows, the management of this data is a critical driver to their success. Every
acquisition brings in new sets of master data to be merged with the existing business. Although
MDM is important to organizations of all sizes, the size of an organization plays a crucial role in how
to approach the implementation of MDM.
Many small businesses do not consider their master data a problem to be concerned with. After
all, the spreadsheets they currently use to manage their Product List work great and the accounting
system is the only location that the chart of accounts needs to exist in. In my experience this is the
easiest and cheapest time to handle the master data management problems. The number of
stakeholders for each dataset is small. The number of systems that rely on each dataset is very
small. This is the time to implement a master data management strategy that can grow with the
business. Each new system implementation will benefit from the single source of all of master data.
Mid-size organizations have a number of dependent systems for each set of master data, so
system integration starts to become important. The number of stakeholders for each silo of master
data is relatively small. These groups may still work in small teams efficiently. Effective master
2. data controls and an owner for each of the master data entities must be defined. These owners,
usually called data stewards, are responsible for managing their domains as new systems are
integrated into the organization.
Large organizations have a number of challenges when implementing a comprehensive MDM
solution. They are large enough that there are several stakeholders for each silo of master data.
Many systems rely on these same models of master data. At this size, coordination of data is a
central concern and requires the input of many different stakeholders.
Conglomerates are the most complex MDM challenges. While they may be smaller in overall size
in term of employees, assets, or revenue than their large organization brethren, the distinguishing
characteristic of these organizations is the breadth of their products offerings or the diversity of
their businesses. Typically these organizations have a diverse offering that makes tracking their
master data very challenging. With a significantly diverse product offering, being mindful of how
the different businesses interact is extremely important. Also, many industries have specific
regulatory hurdles with regards to customers and products that may not be readily known by the
organization as a whole.
Why current systems are ineffective master data applications
In many applications, especially ERP systems, master data is created and stored as a requirement of
these process systems. Some companies may even call the teams that manage the central data for
the ERP systems the master data management group. Although this group is a great place to start
to source the new roles of true master data management, these systems do not provide many of the
features required to properly manage master data for the entire organization.
Some common limitations of these MDM strategies are:
Limited ability to version this master data
Inefficient methods of exporting this data into other applications
Master data is specific to the functions of the system that manages it and doesn’t readily satisfy
the requirements of other applications that need to consume it
Inability to properly store hierarchies or change hierarchies as business requirements change
Limited or no ability to model relationships between different data groups
Limited ability to version this master data
Functional systems require master data to run their specific operations for the organization.
Their chief consideration is the most current general ledger, cost centers, organizational entities,
and products. Companies spend thousands of hours and hundreds of thousands of dollars to reorganize their sales team. Invariably, a large portion of this time and money is spent mapping the
old business units to the new structure.
Inefficient methods of exporting data to other applications
Large, business-wide applications are heavily customized for each organization. These
systems provide limited ability to transfer data out of the master data systems. Most systems have
some export mechanism that resembles a query language with output of text files. The ability to
transform this data with system tools during the export process is very limited if it exists at all. It is
also very difficult to export changes within a specified period of time. Due to the lack of versioning,
it is very unlikely that master data transactions will be available.
Master data is skewed to the functions of the system it is in
3. These systems have been customized to provide tailored processes to the organization. In
the process of customizing these systems, many of the strategies used to customize these processes
revolve around making modifications to the master data stored. These changes may work well for
their intended function, but as we will see ,storing data in a function-dependent manner makes it
less usable to the rest of the organization’s systems.
Inability to properly store hierarchies
Some ERP/CRM solutions tout the ability to store master data hierarchically. In actuality
this is usually managed by placing multiple identifiers into an attached attribute. By giving each
character in this attribute special meaning, a surrogate-derived hierarchy can be formed in any
subscribing reporting engines. This is a messy solution that tends to scale poorly. As a company
grows, each of these character sets can become overextended, creating complex interim solutions.
Changing the hierarchy requires changing the identifiers of all records, which can be prohibitively
expensive.Proper hierarchy storage should allow for both derived-data hierarchy relationships and
arbitrary parent-child relationships.
Limited or no ability to model relationships between different data groups
Many solutions are not designed to allow relationships to be made between two disparate
data groups. Managing products per customer or products per salesperson can be difficult, if not
impossible, as the systems may be working with a small subset of the overall corporate data set.
Watch this space for the second part of this Series
Loving P&C
DC*
Brotherhood of Master Data Management -MDM must be process-Independent
Dears,
Welcome to the second part of this series.Large ERP systems are designed to manage all of the
master data tailored for their system's needs. In this regard they are highly effective, but master
data needs to be stored separately from the processes that use the data. As systems evolve over
time, one of the easiest ways to modify complex system functionality is to modify the data to solve
the problems. Many systems will have multiple customer IDs that map to the same customer to
meet some custom reporting needs.
Another issue with storing master data in a process-oriented system is the need to store
transactional history. As transactions are created, each is tagged with a combination of account,
customer, product, cost center, and so on. These tags must be kept to maintain referential integrity
in the system. These histories are like shackles to your master data, requiring multiple custom
fields to maintain open and closed statuses.
Master data systems should be agnostic to the uses of this data. This approach keeps these
records clean of any attempt to circumvent the programming of a production system. By
eliminating the need to maintain ancient accounts for transactional history, MDM systems can
provide clean representations of each master data set.
Different methods of implementing master data management
A master data management solution can have many different looks. It is very rare to see a large
organization implement a corporate-wide MDM solution in one project. Most of these projects
4. grow organically as different group associated with the project spread the word of the cost and
time savings that MDM enables. There are a number of different factors that contribute to the style
of implementation that is chosen. Some of these factors are:
Level of the organization behind this initiative
Structure of the current organization
Structure of the current functional systems
Complexity of the systems to be integrated
Size of the organization
Degree of internal pain attributed to master data issues
Level of the organization behind the initiative
A true enterprise-wide MDM implementation requires the highest level of an organization
to underwrite the project. If data integration pains are felt at a lower functional level of the
organization, single-dimension MDM solutions are a good fit. Once success is achieved for this one
area, more centralized support for expanding the implementation may be found.
Size of the organization
How many people within the organization are dependent on this master data set? How many
records need to be stored for each data set? These questions help an organization to determine the
type of solution to implement. Once an organization reaches a certain size, implementing an
enterprise-wide MDM solution in one project becomes unfeasible. A phased approach may be more
prudent.
Structure of the current organization
Is the company large and centrally located? Will multiple organizational units need to be
synchronized? Will the internal corporate culture create drag on the implementation of any new
solution? As much as size matters, the current structure of an organization matters more.
Structure of the current functional system
When evaluating each system to integrate into an MDM solution, a number of structural
elements can affect the decision-making process. Are all customers' records reflected in the system
to be integrated? Do multiple records for the same customer provide some functional benefit to the
system? Can these customer records be aggregated in the master data management solution?
Freeing master data from process systems can allow for better data quality.
Complexity of the systems to be integrated
It is important to evaluate the complexity of each system to be integrated. How critical is
the system to the business? What are the best methods to import/export master data from the
system? Will the master data stored have a high correlation factor to other systems within the
organization?
Degree of internal pain attributed to master data issues
As is the case with any IT project, how large a problem the current process creates for the
organization is directly related to the amount of resources that are brought to bear on alleviating
the issue. Without the financial incentive to optimize master data management, many organizations
will choose less costly methods of data integration.
The MDM Cult Series-Part 3-MDM Inception Recipe
Dears,
We have a variety of MDM Application Suites catering to Product, Customer, Site, Activity,
Supplier..etc .Out of this whole bunch two highly successful types of targeted MDM solutions are
customer data integration (CDI)/Customer Hub and product information management (PIM) /
Product Hub solutions.
5. Customer Data Integration / Customer Hub
Customer data integration solutions center around the integration of an organization’s customer
data. These solutions are highly integrated with a company’s CRM and ERP systems. Due to the
nature of combining so many disparate sources of customer data, match merge and duplicate
removal algorithms are critical areas of data integration within the CDI subtype. As this name
implies, most of these solutions are focused on integration and require additional internal
processes to maintain these systems.
PIM) / Product Hub solutions
Product information management solutions provide an organization with product-centric
management. These solutions typically focus on managing, correlating, and merging product data as
bills of materials and online catalogs.
Due to the limited scope of these solutions, many organizations can implement them in a relatively
short time. The limited scope also keeps the number of stakeholders that must reach a consensus to
a minimum. Quick wins and a narrow scope cause many companies to implement these solutions,
ignoring the serious limitations these solutions provide from the perspective of organizational
master data management.Neither of these solutions is designed to be applied to all of the data sets
within the organization. An organization can implement a number of separate solutions to create
coverage of all their possible master data sets. Implementing these multiple solutions reduces the
number of possible integration points but maintains data silos. Cross-dimensional relationships are
difficult, if not impossible, to manage.
Working with IT
It is critical for business owners taking on the challenge of MDM to work well with the
organization's Information Technology group. While many of the MDM management tasks required
for sustained success rely heavily on the business users themselves, management of the
technologies associated with the data integration routines will need to work well with the IT
current processes.
Implementing in a phased approach
Now that we have detailed all the possible techniques to implement master data management
within the organization, let us lay out some efficient ways to implement MDM within an
organization. Each of these processes can provide an organization a reasonable path to grow MDM
through the organization. An organization’s structure will have a significant impact on which
method will be the most feasible.
Single Dimension Build Out
Many organizations come to realize that master data management is needed in their organization
through one significant business driver. Commonly, these issues revolve around only a few distinct
dimensions within the business. These pains within an organization provide the perfect location to
start the MDM process. A few common characteristics that will assist in this approach are:
Single dimension being affected
Low resistance to a new system of entry for this dimension
Minimal additional stakeholders required for complete implementation
Central data management (at least for the dimension in question)
6. For this example, let us discuss a fictitious company called Fabrikam, Inc. This company spends
at least 20 hours per month reconciling changes between seven systems that rely on accountrelated information. New accounts are created by three different groups. The Accounts Payable
group creates a new account for each new supplier. Accounts receivable creates a new account for
each new customer. All other account changes are handled by the financial controller. These
changes and all the corresponding attributes must be propagated to all seven systems. Difficulties
arise because many of these accounts are created a few months before a balance is shown in them.
If a system is not in sync at that time, reports will not balance and it is difficult to determine where
the error is coming from.
Phased Approach – Before
After reviewing their options, the finance and IT departments agreed that this problem was
a commonly recurring issue with the master data within the organization. The IT department was
able to identify at least six places within the organization where critical company master data
needed to be synchronized. While the time and monetary resources were not available to
implement an enterprise-wide MDM solution in the next quarter, it was determined that the
solution provided should have the ability to scale across the organization.
Three months later, the initial implementation was a success. All three account creators
were using the MDM solution to create new accounts. All of the account structures were being
propagated to the seven internal systems. The company spends less than one hour a month
reconciling issues between any of the financial systems. After seeing the initial success of the
finance department's implementation, the sales organization is preparing a project to leverage the
MDM system to manage its active customers. The accounting group will be able to leverage this
data when creating the accounts receivable data as well.
This project is prepared to grow organically throughout the organization, each group taking
advantage of the efficiencies learned in early phases of the MDM implementation. As the project
grows into other areas of the business, it is imperative that clear ownership is determined. These
"data stewards" will be discussed in more detail in the next white paper.
Phased Approach – After
7. Complete enterprise MDM implementations
Complete MDM solutions require the entire lifecycle of the master entities to be managed
from within the master data management solution. Controlling the entry of the master data allows
the enterprise application to proactively manage the quality of the data. Although an enterprise
implementation will be both the system of entry and system of record for all master data entities, it
may still require mapping data to other applications.
It would be naive to think that any company will be capable of getting all of their systems to
use the exact same set of data. Some transformations will still be required to run the process
systems. This does not mean that every defining characteristic of an entity is managed within the
solution. Those defining attributes unique to an organization's system operation should be
managed within the source system where they have relevance. The enterprise solution should
provide a broad range of entry points to be a viable option as the system of entry.
There are four techniques for implementing master data management within an organization.
These methods differ in the amount of control they exert over the master data they manage. All of
these techniques rely on reliable data integration solutions.
In the following section, a number of acronyms will be used within system illustrations. These
acronyms are described below:
Acronym
SOE
SOR
MDIS
IM
BI
Description
Primary point of data entry. This may be direct entry
or through services that update the data in virtual
System of Entry
real time.
Most, if not all systems, receive their data from this
source.
When conflicts arise, this system is
System of Record
considered primary.
Data cleansing and integration processes that
provide automated methods for some of the
Master
Data
Integration following activities: segmentation, aggregation,
Services
transformations, match/merge, and grouping.
To eliminate repeated integration, identity maps
should be used to manage surrogate key
relationships. These may be one-to-one or one-toIdentity Mapping
many.
Technology and applications dedicated to the
Business Intelligence
Long Name
8. analysis and presentation of business information.
Relationship Customer management software.
CRM
Customer
Manager
ERP
DW
Enterprise Resource Planner
Data Warehouse
S1,S2 …
Additional Systems
Software system that serves all areas of a business
enterprise.
Repository of electronically stored data.
These are just placeholders for company-specific
applications that are yet to be defined.
Master Data Registry
In registry implementations, each system remains in control of its own data. All system
data records are mapped in the master data registry. Data maps show the relationships between
the keys of two different systems. These keys can be mapped in two different ways:
One to one: Every record in the main system will have only one corresponding record in the
secondary system.
One to many: Every record in the main system will have one or more corresponding
records in the secondary system.
These mappings provide the data integration applications a reliable way to compare related items.
At any time, different systems can be compared and cleansed. Although this technique provides
important mapping information to the organization, any new items in any system will lead to data
inconsistencies within the solution requiring a very complex data management story.
Figure 1: Master Data Registry
Data aggregation solutions
It is very common for initial MDM implementations to implement this type of watereddown approach. In many cases the applications and systems used for this technique are identical to
the more advanced techniques listed below. The major missing factor in the solution is control. It
is very difficult for an initial MDM project to get all of the necessary stakeholders to relinquish
control of their data to a new product immediately. Another system, usually the most critical
business transaction application, remains the system of record and system of entry. Integration
processes transfer data from this initial source to the MDM application. This master data may be
enhanced by the MDM application itself, but a majority of the important information is still
imported from the more entrenched system. These systems may be more entrenched due to a
number of factors, including importance to the business, amount of time spent cleansing the data,
9. current process, or even perception of value. Data will then be propagated to other systems. Data
controls will be limited as the source system will not be designed to account for any other system’s
requirements.
Despite the limitations inherent with this method, this is actually a good method for bringing quick
wins to an organization with MDM. Many stakeholders are reluctant to give up the security of their
traditional data management processes. By showing early benefits and demonstrating application
reliability, an organization can develop trust in the MDM application as the system of record and
system of entry.
Using this technique for the initial implementation, risk to the mission-critical application can be
mitigated effectively. Less critical applications can begin to source their master data from the MDM
application, solving any integration issues that arise without major ramification to the organization.
Integration processes can be tested and modified in an iterative fashion.
Figure 2: Data Aggregation
System-of-record-only implementations
These implementations give complete control of the master data sets to the MDM
application. Other systems provide the initial data to be imported into the MDM system, but, unlike
the data aggregation solution, the flow of data from this System of Entry is bidirectional. New
records are transferred into the MDM application for integration. Any discrepancies in the data
defer to the MDM application, which is the system of record.
These implementations still require a degree of data integration and ongoing cleansing as
elements may come from both the source system and the MDM application. Also, many times this
system of entry only has the ability to detect data issues directly related to the initial use. For
instance, any customer information that is not stored in the CRM solution will not be available to
determine complete data quality.
Figure 3: SOR only (Hub)
10. Complete enterprise MDM implementations
Complete MDM solutions require the entire lifecycle of the master entities to be managed
from within the master data management solution. Controlling the entry of the master data allows
the enterprise application to proactively manage the quality of the data. Although an enterprise
implementation will be both the system of entry and system of record for all master data entities, it
may still require mapping data to other applications.
It would be naive to think that any company will be capable of getting all of their systems to
use the exact same set of data. Some transformations will still be required to run the process
systems. This does not mean that every defining characteristic of an entity is managed within the
solution. Those defining attributes unique to an organization's system operation should be
managed within the source system where they have relevance. The enterprise solution should
provide a broad range of entry points to be a viable option as the system of entry.
Figure 4: Enterprise Solution