Breaking News

Kalido

 

 

MDM Hub Architecture - Welcome to Utopia!

Here in Utopia, we never have arguments about whose figures are right.  We never need to reconcile mismatches in numbers.  We never cold-call our own customers trying to sell them products they already have.  We have a master data hub!

 

Our hub is the one and only place that we store the undisputed master of our shared data.  It has a service wrapper, so any system can request current, historic or even planned future states of the data on demand.

 

Our systems access the hub to get master data when they need it, and don’t even store their own copies of the data.  If any master data changes, the change happens in the hub so the next time any system accesses the hub it gets the new version of the data.  Our IT organization has the hub protected like Fort Knox to ensure data only gets out to authorized consumers, and it runs on the latest infrastructure that makes sure it carries on running even if our data center gets taken out by a hurricane.

 

For data that can be sensitive, or where changes could have a potentially serious knock-on effect on the parts of the business that use it, we have a governance program that ensures business managers carry out due diligence on the data changes and authorize them before they are made in the hub.

 

Ahhhhh….life is good!

 

Wake up!

Like coming round from a daydream, we are jolted back into the reality of today.  Our legacy systems, custom systems and purchased packages all store data that ought to be the same, but isn’t.  Our business intelligence gives us different answers depending on which way we ask the question, and we spend hours massaging and reconciling data in spreadsheets to get numbers we half believe in.

 

Our systems often deliver the wrong product to the wrong place.  Customers update their personal details, but we keep using the old details for months, making us appear unprofessional and uncaring.  And every time we try to do something about it, we run into what seem to be insurmountable difficulties.  We can’t agree on definitions and business rules across different business units.  We can’t get changes made to packages and legacy systems in any reasonable timeframe.  Our data is such poor quality that in some cases it seems it would be simpler to throw it all out and start over.

 

Utopia seems like such a long way off.

 

Transition strategies

How can we get a grip on the current state of our data and make progress toward a Utopia? 

 

Sorry to tell you, there is no Silver Bullet.  However, consider the following three broad strategies as alternative ways to set off on the journey towards Utopia:

 

Utopia bit by bit: One subject at a time

In this approach we will build an operational hub “one bite of data at a time.”  The bites can be quite small (for example customer number, name, billing address, main delivery address and credit rating) or more ambitious (for example “Customer data first”).

 

Pros:

-          Starting small means that you discover problems early and in an environment that has less overall impact. 

-          You can build skills and experience and hone your approach as you go

-          It is less risky than a “big bang” approach (though anything that requires open heart surgery to operational applications is always risky)

 

Cons:

-          It takes longer to get there.  The overhead of managing a program with many small increments are higher than a program which bites off a significant chunk

-          It can mean significant rework to (small) increments completed earlier which can be tricky

-          It may disappoint people expecting very quick returns to justify the MDM investments

 

So it is a balance between the risk of a larger bite and the overhead, and rework of smaller bites.

 

Utopia bit by bit: One Process at a time

This is a variant of the first approach oriented towards all the master data needs of a single business process at a time.  It fits well with process improvement/process re-engineering programs.

 

Pros:

-          Limits/isolates change to a single process (which are happening anyway if a process improvement program is underway)

-          Usually has a clear single point of business management to make calls on difficult decisions (the process owner).

 

Cons:

-          Likely to involve re-work when the second and subsequent business processes are tackled

-          The project may be pulled toward the needs of a single business process and make it difficult to develop a true enterprise-wide hub over time.

 

Analytic first

This approach focuses on master data for business intelligence and data warehouse systems first.  It isolates inconsistent data in operational systems from the analytical layer.  This ensures that a consistent set of data goes into BI systems improving the quality of data coming out of such systems, which in turn results in better decision making.

 

Pros:

-          Data integration skills and experience already exist in the data warehouse/BI community

-          Technical pressures on a BI system are less imposing than on operational systems

-          The approach is less risky overall because operational systems do not need to be modified.

-          First results can be achieved quickly

-          The initiative can afford to be enterprise wide (even if it still only tackles a subset of the data)

-          Once the central repository is populated with harmonized data, operational systems can be gradually bought in line while the BI layer remains consistent.

 

Cons:

-          Data governance is more demanding as much of the data required for analytical purposes is a matter for human opinion and judgment (e.g. classifications and categories that can be used for analysis)

-          Data quality efforts will need to be applied continuously to resolve inconsistencies on new data created in operational systems until the hub can take over as the central point of changes.

 

Whichever route you take to get to Utopia, you should bear in mind that Utopia needs to satisfy operational and analytical requirements for MDM — the destination is the same, it’s just the route to get there that is different.

 

Visit www.kalido.com.

 

 

admin @ July 30, 2008