Spending time at the MDM/DG Summit in NYC this week demonstrated the wide spectrum of MDM implementations and stories out in the market.  It certainly coincides with our upcoming MDM inquriry analysis where:

  • Big data is influencing MDM strategies and plans
  • Moving from MDM silos to enterprise MDM hubs
  • Linking MDM to business outcomes and initiatives
  • Cloud, cloud, cloud

Taking a step back from this, these items beg the question – how do you apply MDM within this new data ecosystem?

During my keynote I tried to expand our perspective on MDM to be a hub for context in customer experience – sitting between systems of record and systems of engagement to translate, manage and evolve dynamically the full fidelity of customer identity through interactions directly or as viewed through indirect business processes and supporting activities.  For example, what can you learn about an air traveler by analyzing baggage handling data?  What can be learned about annonymous visitors to your website beyond the activity tracked on site – what did they look at prior to coming? what device did they use? where are they located? do you already have past behavior stored in a cookie? This opens up the customer master beyond the 25 – 250 data elements we might include today to potentially thousands or customer markers that define identies.  All this metadata is master data.

Now that I blew your mind, is this even possible?  Oh yeah baby!

Let's talk graph databases.  Back in 2013 Facebook launched the Facebook Graph Search beta.  This took the idea of six degrees of separation and let you navigate through these connections based on your interests and natural language request.  Behind the scene, the database is not structured around a logical structure of tables and data elements.  It is based on classifications and relationships.  This provides dimensionality way beyond the regidity of relational database repositories and requires little to no translation of how we view people based on the data we have (system, tables, user interface, etc).  It's semantic – the human view.

MDM is not a data integration tool!  (say this loud and clear so I can hear you – then say it again)  Even as MDM is founded on the premise of managing simple to complex data models, the fact that the repository is typically XML or relational and it is designed as a system hub turns it into a pretty powerful data integration tool.  And thus, this is how we have typically implemented it.  

But, what if you change the underlying repository to a graph? It immediately changes the mindset and strategy of MDM from systems to views.  Much more intuitive, analytic, and intelligent about our master data.  And this is what innovative MDM companies are doing – using a graph db repository (ie. Pitney Bowes Spectrum MDM).  And, still other innovative organizations are saying, we can build this on our own by leveraging a graph db (good confirmation and examples of this with Neo Technology).  And, you have data profiling and discovery tools like Global IDs helping you identify and build a graph of your data (they OEM Neo4j from Neo Technology and use the open source graph db Titan). 

There are still challenges:  

  • Scalability can be limited – to overcome this companies are sitting their graph dbs on top of triple-stores to overcome this.  
  • Semantic and graph skills are limited – analytic teams are being tapped to support data modelers and MDM developers.  You many need to invest in semantic discovery and modeling tools.
  • In-Graph MDM tools are still new – start with well known domains and business/engagement processes and scale from there to avoid too much innovation at once.

What we need to do is demand from vendors capabilities that keep pace with the complexity of a contextual, dimensional and dynamic ecosystem that operates at extreme scale.  Emphasis on processing speeds, integration points with different platforms and applications, and improved development and governance.