Your Master Data Management Tool Should Not Dictate Your Business Strategy
While at Siperian (acquired by and now offered as Informatica MDM) our biggest competitor was Initiate Systems (acquired by and now sold by IBM). One of the key differences between the two master data management (MDM) offerings lay in how “golden records” also known as the single/best version of the truth (BVT) were assembled. Informatica MDM/Siperian matched and merged into a consolidated, persisted BVT, while IBM/Initiate matched and linked into a registry style virtual BVT. Which camp you were in, depended upon what your business was trying to achieve at the time, and what governance processes you believed were more suitable for your organization. Both systems had their advantages as well as their drawbacks. A persisted BVT for example would not reflect real-time changes in the data, while a virtual view required instantiation before it could be shared with downstream analytics and data warehouses, and it could yield inconsistent results.
These legacy master data management architectures were limited by the technologies that were available 10 years ago. Today, the demarcation between operational and analytical is blurring. Big data technologies, increasing compute capacity and cloud deployments are shattering traditional paradigms. Another example of this can be seen with ETL only tools now supporting real-time integration, and EAI tool vendors now offering batch integration.
Our belief is that enterprises should not have to distinguish or settle for persisted or virtual views. Both are equally important, with quality and consistency the overriding foundation. Multi-domain profiles and multi-dimensional relationships should be stored in a commercial graph when data is matched, merged or modified with a full history and audit trail. A snapshot or persisted view can be delivered at any point in time and exported as needed. As long as the architecture is a modern one, as much compute capacity can be applied to provide the right and consistent view through data-driven business applications in context with the request of the end user. Data can be cleansed at every point of entry (UI, batch, APIs), regardless of which source it comes from. When data is searched, queried or exported, views can be dynamically assembled and delivered in real-time using in memory technologies.
Additionally, since most enterprises today need business agility and must satisfy a variety of compliance requirements, they cannot be shackled with a single model of governance. They should not have to choose between centralized or decentralized governance strategies. Modern data management platforms and applications must support mixed modes because attributes that need to be managed can come from multiple areas at any time, including: CRM and ERP systems, third party data sources, process-oriented collaborated curation, and even inline analytic processes such as demographic and psychographic segmentation.
@Reltio we believe the time is right for enterprise data-driven applications with modern data management that is free from the constraints of legacy MDM offerings. Not only is it expensive from an ongoing maintenance and management perspective, if it is dictating or keeping your hands tied when it comes to your business strategy you are losing much more than you realize.