Author Lochan Narvekar, © 2015 Ref# MDM0001-R01
What is
Master Data Management?
Master Data
Management is really a business problem that is Age-Old. How do I get best out
of my business processes?
But to answer
this question one should get the underlying data as well in order! I like to
quote, “Quality processes need quality data, but also, vice-versa”. By data I refer to transactional as well as
Master data. By consequence, transaction data will depend heavily on the base
master data. If there is confusion on master data, how I can guarantee my
transaction data to my business? As you can see, this is really an old problem
and hence there is multitude of solutions over last so many years. In this
article, I will focus on MDM principles I have developed after years of work in
this area. There is always scope for improvement, but objective of this article
is to set formal base line for understanding MDM scope.
Common MDM Concepts
Let’s look at
each area,
Ownership
After we
scope the MDM effort, we will arrive at a set of attribute groups that are
going to be under MDM control. It’s essential to set Ownership of these
attributes. Meaning which Roles are accountable for the values in these attributes?
This helps in getting data cleansed before the Go Live. Also, it helps after go
live in making these Roles accountable for accuracy of these attributes.
Common
Definitions & Classifications
An attribute
called one thing in one system should not be called as something else in other
system. This seems like a common sense, but where each system has different
organic growth, different owners, this is bound to happen without MDM understanding.
MDM helps us tackle this issue from top down.
We also need
to look at the classification for data element as it drives the attributes
needed for that data element. Each classification has different required
attributes. As much as possible we need to match the classification across
business systems.
E.g. Products
are classified as product types in an Engineering system. We must strive to communize the same product
types in Marketing system with same names. There might be more classifications
in Marketing and more attributes, but common ones need common names
Logical Data Models are
great aid in documenting anomalies and flow of data between systems.
Data
Cleansing
Data
cleansing has 2 phases to it. One, before Go Live and another, after Go Live.
It has surprised me to see that often folks miss on the latter. Imagine
cleaning the data, transforming it and after Go Live, it again gets dirty. This problem can be solved with following
approach.
First, get a
good data cleansing tool and implement it alongside the MDM tool. If the data
is small then, manual effort is good enough.
Second, we
need to put checks in place to not allow creation of bad data going forward.
Treatment here is different for internal vs. external applications.
Third, we
have to devise a way to clean the data on on-going basis with Data Stewards,
and reports.
MDM
Systems and Integrations
This is the
most important pillar. Management needs to study the current application set
and decide if the organization needs and/or is ready for a new MDM application.
Lot can be achieved without implementing a new tool, agreed! But this needs a strong adherence to MDM
principles, and a good consultant. I have often reconfigured the existing
toolset and with help of MDM principles tightened the business area in
question. But, having a specialized tool for especially if the scope is big
enough generates good ROI.
Many
companies have Enterprise Integration platforms or Enterprise Bus. They play
key role in MDM projects tying together the MDM tool and/or all the systems
through MDM principles. Newer platforms even offer embedded Business Process
Managers to weave the tools together. Remember to look at all of this through
MDM lenses as described here.
Please read article “ROI
on PIM: How to build a Business Case Ref# PIM0004-R01” as an example of ROI
justification. Similar ideas can be applied to CDI and other MDM areas.
Business
Process Change
As I wrote
above, “Quality processes need quality data, but also, vice-versa”. What this means to me is, there is a business
and IT partnership in MDM success. What good is the data if the processes are
allowing data to change uncontrollably? This change is not drastic. Few checks
and balances in the process based on MDM principles to make sure that data
stays good.
Please read article
“PIM: A classic High-Tech case study “Ref# PIM0003-R01” for a good example of business
process change.
Data
Governance
This is another
important pillar! Good data is only going to stay good, if the good data
governance is put in place.
This begins
with the Attribute Owner. Who is ultimate authority on this attribute (perhaps
more adequately, this business sub-function)?
Then, who can
see the attribute or group of attributes? Who can change them? Who needs to be
notified of the change? This is most important topic and one needs to use
templates to aid in this. Who needs to subscribe to the change? What systems
need this data? All of these questions
need to be addressed.
Metrics
and Reporting
We need to
put together right metrics so that after go live we can measure the performance
of our MDM implementation. And we need
reports on top of these metrics.
About the
Author: About the Author: Lochan started his career in R&D
architecting PLM/MDM software. Moved to Consulting in 2005, and has been
successfully consulting for many years in the PLM, PIM and MDM area. He can be
contacted Lochan@gmail.com.
This is copyrighted material; © 2015 Lochan Narvekar.
