Next up was a session from some folks at ASG talking about business metadata. They started by discussing the metadata audience and how it is changing as the syntactic and semantic richness of metadata increases. Initially there was a focus on consistent definitions for, say, COBOL copy books. Gradually expanded out to DBAs, Data Architects then to Information Professionals (Business Intelligence, Data Warehouse, Content Management etc) today. In the future knowledge workers themselves will need to work with metadata. Metadata is needed as it provides information context that supports effective use of information assets. They see metadata as part of closing the IT expectations gap – the gap between what is expected from IT and what is delivered.
Business metadata is the metadata that is required for workers at a company to successfully interact with customers and perform their job. This kind of metadata needs to be preserved as part of the institutional knowledge and is focused on accelerated business decision making. This business metadata mediates between business activities and IT activities. As business users composite their own mashups and processes they need more and more operational details in their metadata. Business metadata can be broken down ito various categories:
To align business information were it intersects with technology
To define publication
To ensure compliance and collaboration on established terms
- Structured and Unstructured
Business metadata is operational
Business metadata can come from a range of sources. Including:
- Business Service Operation
Transaction and audit information generated by operations
Ideally a byproduct of the SDLC or from structured discovery
- Human – extracting it from experts
It really needs to be a byproduct of development and operations. Scheduled and manual approaches are OK, but it really needs to be produced as other activities go on. For instance, as data is extracted from an operational data store and loaded into a data warehouse, need to capture metadata on data lineage, business rules used to transform etc. The integration of the various tools is lacking, however, so there remains a fair amount of manual work. Business metadata must also be communicated so that users can rely on it, for instance, when developing in Excel. Sharing this information must become part of normal life. It must evolve from “reported” to “by query” (current state) to “on demand” and, presumably, to “invisible” where it just happens.
They are clearly starting to discuss business rules in this context, as business metadata, and that’s a good thing. I think multiple layers of rules become important in this context – both high-level “source” rules and operational “production” rules that actually execute. I also think that elements like data mining, predictive analytics, adaptive control approaches also need to be reflected in metadata but I don’t see this happening much yet. Having some focus on decisions will also help, I think, as knowing what decision is at issue helps focus the metadata needed.
Lunch then my session.