Masters of Their Data Domains
- By Alan R. Earls
The Big Idea
PLANNING TO DEVELOP
- Enterprise apps increasingly support the services approach to development. ERPs are no exception, featuring data schemas, object sets, APIs, sticky middleware and GUIs, among other tools.
- Some enterprises are using these industrialstrength ERP systems as development platforms for service-oriented architectures.
- New ERPs are much easer to work with, say systems integrators, but if you're using them as app dev platforms, you'll need to carry a heavy tool bag.
Picture the old, painfully funny Abbott and Costello
“Who’s on first” routine in which the comedians drive
each other to distraction by misinterpreting the
names of players in a baseball game. When many companies
try to view crucial customer or product data, misinterpreting
basic information is a common problem. Enterprise apps
often provide different versions of what is supposed to be the
same data—versions which are incompatible in structure, definition
and content. Only the results aren’t funny. They are costly
and affect competitiveness.
That’s where the concept of master data management comes
in, and not surprisingly, MDM is defined in different ways. IDC
describes MDM as a set of processes to create and maintain a single
view of products, customers, accounts or locations, through
a physical or logical hub across the enterprise or, in some cases,
across enterprises. Although the concept is simple, getting there
So vendors and consultants have been gearing up with technology—some of it old, some of it borrowed and reworked from
other data-handling tasks, and some of it new—to give orgs a
chance to get data right once and for all. Vendors vying for this
new market range from startups and niche vendors such as Kalido
to the venerable Big Blue.
IDC expects the annual MDM market to reach $10.4 billion
worldwide by 2009.
Businesses are paying more attention to MDM for several reasons,
according to IDC research. Many enterprises must comply with federal mandates. Companies are also
looking at MDM to improve supply-chain
A single version of truth
"Maintaining a single view of products, customers, locations and financial accounts is a significant challenge," says Henry Morris, group VP and GM for IDC's application strategies research.
"At its heart, master data management involves changing, consolidating and rationalizing core processes such as new product introductions or customer or patient/citizen/supplier registration and ongoing management,” Morris says.
Established terms such as product information management, customer data integration and financial consolidation are attempts to provide a single view of one type of master data, an effort IDC calls applied MDM. Some software of this type came on the market in the 1980s. What's new, Morris says, is the emergence of MDM infrastructure software that is purpose-built to support all types of data, using a combination of techniques from data and content technologies.
Morris reasons there are three types of current MDM scenarios, reflecting the diverse set of projects now being undertaken:
- Management Reporting Scenario, which reconciles, rationalizes or organizes master data at the hub to drive reporting for compliance and business performance management (BPM).
- Data Synchronization Scenario, which builds consistency in operations by synchronizing master data from the hub back to local systems.
- Single Point of Origination Scenario, which requires that all changes to master data originate at the hub for real-time consistency in operations. As the scenarios become more challenging, increasing levels of services are needed.
—Alan R. Earls
Master plan to master data
MDM is fundamentally a technology
and a business issue. “In order to tackle
it effectively, companies must make
some definitive decisions about what
they want to define as master data—for
customers, parts, bills of material and so
on,” says Josh Greenbaum, an analyst at
The business issues are especially
complicated because they involve centralization
of authority and a level of
command and control that most companies
are unable to implement—sometimes
for good reasons, Greenbaum observes.
Many companies prefer a more
decentralized business model, but that
approach carries a penalty in terms of the
additional costs incurred for customer
and parts acquisition.
If a Global 1000 company had access
to a single master parts list as a result of
implementing MDM, leverage in acquisition
would be a huge benefit. Still, it’s
a balancing act. “You don’t want to start
an organizational revolution just to have
a single parts list,” Greenbaum says.
Sure, getting 100-percent control over
data—master data or any kind of data—is probably impossible. “The real obvious
pain points, like customer data, are
something everyone can afford to tackle,”
People who need people
Andrew White at Gartner believes people,
not technology, are the primary challenge.
That doesn’t mean the technical
challenges are trivial. MDM will have a
tremendous impact on the way apps are
developed. “Until now, the assumption
has been that each app owns its data, but
the MDM movement is based on something
very different,” he says.
What’s the state of MDM adoption
today? Companies have known all along
they had a problem; what is new is the recent
wave of solutions, which Gartner
lumps under a concept called enterprise
So far, White says, it’s the big companies
that are putting money and thinking
into ways to solve MDM, and these orgs
are begging vendors to help. But vendors
are only just beginning to realize that in a
world of SOA, interoperability is going to
“In order for service-oriented architectures
to work, there needs to be a
whole layer of master data reconciliation,”
White says. Vendors now realize, according
to White, if they don’t fix this issue,
Web master data
Despite the lack of agreement about exactly
how MDM will happen, what it is
and where it is going, practitioners are
helping their customers tackle MDM
problems today from a number of different
angles. Two companies with comprehensive
MDM visions, according to
analysts, are Kalido, a spinoff of petroleum
giant Shell, and IBM.
Kalido sells two apps: one focused on
data warehousing and the other on
MDM. The Kalido 8M: MDM product
manages and controls multiple types of
master data, not just product or customer
data. It can take data from any
number of systems then support analysis
that will, for example, accurately
show which items are the most profitable.
This type of functionality is a result
of the company’s work in BI before
it ventured into MDM, says Cliff Longman,
Kalido’s customers are deploying
MDM to manage common business definitions
for master data within a workflow-driven, Web-based repository. The
repository is then used for corporate performance
reporting and analysis to increase
its consistency and accuracy, reduce
errors and obviate the need for
standardization of the underlying operational
Master data can be published and distributed
to business people as well as operational
systems in order to share the
common data definitions more widely. A
history of changes made to master data
over time is kept to support trend analysis
and audit trails for compliance and
“If you look at our biggest soft spot,
it is B2B with complex customers,”
Longman says. In B2B environments,
customers are frequently changing customers,
partners, and products and operate
using complex IT infrastructures.
A key design feature for Kalido is using
just one table to store all data: “If you
change the structure of the data or add
new levels and hierarchy to adapt to a
change in the business, the programs that
access the data will still function,” Longman
explains. In effect, he adds, this approach
separates the business model from
the underlying database schema.
Governance is a big piece
Big Blue is also eyeing the master data
problem—and the market opportunity.
IBM’s WebSphere DataStage TX enables
enterprises to combine data, apps,
and messaging system content and deliver
the resulting data to any app, business
process or end user that needs it.
The software, which is built on an SOA,
allows the data integration environment
to be deployed in an open architecture.
Because it supports e-business standards,
WebSphere DataStage TXcan integrate
business data (XML and non-XML) regardless of data volumes—for
example, a HIPAA claim that has an
XML claim attachment embedded inside
a non-XML document—while ensuring
data quality, according to IBM.
WebSphere DataStage TXprovides “industry
packs” to simplify the task of integrating
data from disparate sources and
transforming it into a format that meets industry-
specific standards such as SWIFT,
HIPAAand EDI. The software also supports
integration of data from widely used
enterprise apps from vendors such as SAP,
Siebel, Oracle and PeopleSoft.
Technology isn’t the whole battle, says
Craig Jett, MDM strategy and marketing
executive at IBM: “Abig piece is governance
and how you manage your data.”
MDM is accomplished through middleware
and services that support data
management and put data within the
right business context for a particular domain.
Data process control through an
SOA is a key component of MDM, says
Jett. “Ultimately, we lump MDM into
data governance, stewardship and change
management,” he says.
No need to go global
In a market as potentially large and diverse
as MDM, not every vendor is aiming to
provide a global solution to everyone’s
problem. Managed services provider Liaison
Technologies offers Enterprise Content
Director, a hosted product information
management (PIM) solution, which
focuses only on product information.
Hosting solutions can cut implementation
time to weeks rather than months and provide
a rapid ROI, asserts Bob Renner,
CEO of Liaison Technologies.
According to Renner, Enterprise Content
Director applies adaptive intelligence
techniques and data manipulation
rules to pull information from data
pools, using industry standards such as
UCCnet, papiNet, CIDX and XML to
ensure accurate data synchronization.
Barbara Mowry, president and CEO of
Silver Creek Systems, says her company
is focusing on “semantic integration” to
deliver MDM’s missing link. What’s
needed, she says, is a semantic-based approach
that ignores data patterns and
identifies real meaning. Once an item is
properly understood, it can be standardized,
localized and enriched according to
the needs of the company and the system.
Silver Creek’s DataLens System, architected
as an SOA, employs AI, semantic
modeling and expert system techniques
to achieve high recognition rates and
reusability, and it does it on the fly and
with easy scalability, according to Mowry.
Ultimately, any business pursuing
MDM is still battling with the oldest IT
problem in the world: “Garbage in,
garbage out.” But unlike the old days, it’s
no longer adequate to merely complain
about the problem, you have to fix it.
ILLUSTRATION BY MICHAEL MORGENSTERN