Question: Should the metadata layer that sits between the data marts and the query/reporting tools be owned by the business or IT and what are the pros and cons of each approach?
Sid Adelman’s Answer: It depends on what is meant by “owned.” IT should provide the metadata repository along with the rules about who can add, update and delete from the repository and IT has responsibility for the technical metadata. The business has responsibility for the creation and maintenance of the business metadata including business definitions, business rules, security and valid values and this business metadata would be housed in the metadata repository that is implemented and supported by IT.
Tom Haughey’s Answer: Particularly from a data warehouse (DW) point of view, there are several key points about metadata:
§ It is vital to the DW;
§ It must be created progressively as the data model is being created;
§ It must include all relevant components; and
§ It must be published in a practical medium.
I feel it is important to understand each of these points in order to properly understand where metadata fits in the spectrum of things.
As you know, a data model has three main components: the model diagram with its entities, attributes and relationships; metadata, or definitions of everything in the model; and, finally, supplementary business rules that cannot be expressed in the structure of the data model.
Metadata for the DW should include definitions of the objects with some sample values, identification of the sources (where did the data come from), description of the business and technical rule used to transform it, timing of the transformation (e.g., data on the last day of the month is not the same as data on the last Friday on the month), appraisal of its quality, standard queries, etc. Metadata is crucial to a successful data model and is indispensable for a DW. Remember, there are both logical and physical data models. There will be metadata for each of them. The physical metadata may be slightly different than logical metadata, because the physical data model may have some differences, mostly due to design differences, trade-offs and additions. In all of this, it is my belief that the business is the best agent to provide the metadata. Even so, it often falls on the data model developer to provide the first cut of the metadata. A convenient way is for the data modeler to gather as many relevant sources as possible, and use them to create the first version of the metadata. This metadata is then passed to relevant business and systems people for them to validate, correct and complete the metadata provided.
Be the first to comment on this post using the section below.