Semantic Web + Master Data Management = Huge Innovation Oportunities?
Some weeks ago, I participated in a workshop about the Semantic Web at the FU Berlin, which was quite interesting. During that day I begun to ask myself how it could be possible to connect the systems used for Managing Business Master Data with the idea of the semantic web to build a corporate data web which solves all the integration and retrieval problems Data Warehouses, OLAP, ETL (…) EAI, SOA, and Idenity Resolution Systems shall solve.
The more I thought about that idea the more I was confident that this will be a huge trend within the next 5 to 10 years. With RDF, Tripple Stores and their brothers and sisters enterprises will be able to connect the data within silos with the data of other silos. They will be able to retrieve external data and connect it on the model level by just extending and or transforming the underlying moder graphs. This idea, as any other idea isn’t new, as this link of IBM Research shows, but its still the very beginning of a fundamental swith in thinking about and handling corporate data: Humans and Maschines will be able to walk the graphs of data, identify relations between unknown entities and generate hypotheses and assumptions about the real world facts by just allowing the data to be connected by meaning and relationships; its just like it is in the real world.
To complete this post, I wanna give you an impressive List of videos and podcasts about the semantic web. Happy Watching:
http://www.semanticfocus.com/blog/entry/title/302-semantic-web-videos-and-podcasts/
The problem is that semantic technologies bring to few benefits to MDM itself and to the way these Master Data are managed. Examples shown on IBM website do not show a sufficient effort
Ongoing works on MDM is already focused on de-sillowing, bringing data governance within the organization, creating taxonomies and dictionaries to describe Master Data…
An other problem is that most companies already invested a lot to change their organization to handle Master Data accross silos, setup MDM repositories, and applied the required huge amount of data quality processing to centralite and manage these data. As IBM suggest it, The best way for SWT to come with MDM is that Semantic Technologies should be wrapped around each repository to make it SWT-Ready.
In fact, Semantic Technologies may more benefits from MDM to put a foot within enterprises than the other way round: I think about enterprise common data models (Product, Customers, etc.), I think about taxonomies and dictionaries, and related processes and organization to maintain them.
MDM has a business argument: Bad quality of data makes your customer unhappy, and makes you losing money.
What is the business argument of SWT? “Solving all the integration and retrieval issues” is meaningless to business managers - companies are not reluctant to continue spending their money using old-fashionned P2P ETL and are quite happy with that.
One possible scenario is that an other critical business issues like risk management or BI sees a huge benefit of using semantic technologies, like integrating external data (I think about data.gov, or such initiative). And as far as I know, we are to far away from having an enterprise global ontology repository to allow reasonning within an acceptable amount of time, but this may help Risk Management and BI.
Once such a business driver found, I see no problem for the companies to expose one by one their Data repository to be SWT-Ready… as long as they remain secure, accessible by authorized persons only (but SWT may not be the right layer to deal with that… or?).
Killer Business Argument to get in Enteprises -> Success -> Exposure of more and more data repositories -> More SWT-based applications ->… and then, like you said, huge innovation opportunites.
Thanks for your comment Oliver. I completely agree with your points. Just think about storing all static, or semi-static entities such as addresses, company names or person names in nonrelational (key-value stores, column based stores…) data repositories that have functional, service oriented facades that deliver URIs for each entity. To connect / relate entities we use triple stores and application modules with rdf abilities… connected entities become “living” entities (firstname, lastname –> address –> date = living person). When we understand how to exceed those inner organizational models with external commercial knowledge such as Registration, Solvency or Financial Data… there might be large improvements in how we handle data.
Such a lean and easy to access MDM would also offer new possibilities for smaller and more agile consulting companies who are dependent on data.
From my experience the IT overhead (connect database, transform xml, import flat files) for establishing BI, analysis or search components in companies (even small and medium size) is far too much.