Wednesday, July 19, 2006

Tech: Master Data Management

When someone recently asked me what I thought of Master Data Management, I said Hallelujah!

In actual fact, implementation of MDM is frequently incomplete, often incoherent. All enterprises need to maintain a consistent view of their information across all their businesses. To my mind – and many others - this mean keeping tight control of a central data dictionary, that is referenced by all applications and databases. This necessarily includes maintaining different versions of the same data over time – for example, a development system may be working with a new version of that data – but a central reference point is key.

Surprisingly, this doesn’t happen often enough. Part of the problem lies with the need for disparate vendor products (databases, BI systems, etc) to be able to (forced to) synchronise with what will usually be an external reference source.

It’s been on people’s minds for a while (see, for example, these articles from Intelligent Enterprise and Baseline Magazine), but so far I have not seen a perfect implementation. In fact, where I have seen attempts to control data formats and definitions, it’s usually been through manual processes such as source control systems. That’s simple, but not always quite enough to mandate a consistent data view. Software vendors often present products that aim for the ideal – see for example, Hyperion’s MDM . And according to Baseline, Gartner finds favour with its presence in the ETL space (eg IBM/Ascential’s DataStage, and Informatica). However, as DMReview’s take on the matter suggests, it’s a process to work on, and you may never reach a perfect endpoint.

Yet I’m always on the lookout for ideas and products that work towards this ideal.

Comments welcome.

20-Jul-06 Update: As a succinct illustration of how current this issue is, I received an email less than 24 hours after this post, inviting me to participate in a survey on Master Data Management.

26-Jul-06 Update: See also a posting on Master Metadata. This is intended to refer to enterprise-level data about the data, processes, repositories, toolsets even. In effect, data dictionaries of the enterprise. Laudable - an enterprise really needs to be on top of the disparate IT systems - but I expect successful implementations would be rather rare. You'd see the typical problems that aggregate at the enterprise-level, such as speed of change, political manouvring, etc.

No comments: