Governing metadata can be easy to understand, if you start by addressing it at three key levels:
- Semantic (Vocabulary)
- Business metadata (Dictionary)
- Technical metadata (Data)
Robert Seiner, President and Principal of KIK Consulting and Educational Services, brought that message to attendees of DATAVERSITY’s® October Data Governance webinar. As a specialist in Non-Invasive Data Governance™, Seiner discussed the interconnection that exists across all these levels. He examined how the core of successful Data Management and Data Governance efforts lies with developing an architecture that addresses each component individually and, by extension, holistically, fostering transparency, supportiveness, and collaboration. “The three layers stand alone but they also have a lot of things that relate to each other,” Seiner explained.
The value of the semantic layer starts with its ability to decrease frustration among business managers who find that the answers to the questions they ask differ depending on whom they ask. That’s because in too many cases, the same terms are defined differently in different parts of an organization. “The semantic layer is important to helping improve understanding and cooperation around the data in an organization,” Seiner said. It provides the vocabulary – the true business terms – that management wants staff to use in talking to each other and to others, so that everyone is on the same page. Then, “when someone asks a question about something in the organization, we all have the same understanding of what that something is,” he said – as well as what domains or subject areas it relates to.
The business metadata layer – that is, the data dictionary in various environments – encompasses the business descriptions of data in these systems. Think in the context of the definitions for the physical data that people will access in business terms – the name of the piece or element of data and its definition within a particular database or other system, including the domain or sub-domain with which it is associated. For instance, in a manufacturing business people may want to look for information about raw materials, and in all the different resources where such information is available, they have to be able to relate the vocabulary to each resource’s own data dictionary. Seiner calls it “relating terms to business data through metadata.”
The technical metadata layer is the data layer itself – the columns, databases, and tables, and the lineage of and relationships between these pieces. How this layer may inter-relate with another layer is visible in this example: If an organization has true semantic vocabularies, employees may be able to start with a business term in them and dig down to see what technical metadata is available for that term to help answer questions around it. Seeing the rows, columns, views and databases that have that associated information “would be a big benefit to the business people in the organization,” he said.
Start the Journey
Seiner wants to take the understanding of metadata – which most data pros equate to being data about data – to a higher level in pursuit of better Data Governance. “It is data collected in tools of information technology that improve the business and technical understanding of data and data-related assets,” he said. Whether that tool is a spreadsheet or data modeling system, “in order for something to become metadata it must be documented and recorded somewhere,” he said. That way, it is accessible and once accessible, usable and enforceable.
“At the end of the day, when we put Data Governance in place, we execute authority over the management of data,” he says. But he adds that this must be done in a non-threatening way, “so that people adapt and absorb it into day to day culture.”
How to move forward in this direction? For one thing, start by asking the businesspeople in the organization about what kind of information they could use, if it could be made available, that would help them make better decisions based on data. “It’s important to focus on the business aspect of metadata,” he said, “and that means involving stakeholders.” Not that data professionals should necessarily use the word “metadata” in any of these conversations, he advised, as that word tends to produce blank stares from many business users. Instead, “ask them what information they can use about data to do their job more efficiently and effectively,” and the result will be like opening a spigot that will help data pros work intelligently.
It’s also useful to look at scope and levels of complexity the organization is willing to take on. Is there a specific problem to solve, a long-term vision, both? Come to agreement on that, perhaps by aiming at some low hanging fruit first and then looking at a longer term vision of how Metadata Governance fits into the overall culture and desire to manage data as an asset. Unfortunately, without written consent in advance, perhaps given by a steering committee, it can be difficult to assure that the resources always will be available to be certain that your organization can continue governing metadata as it needs to be governed, he cautioned. People will need to be available on an on-going basis, for instance, to pull business terms and terminology for an organization’s expanding business vocabulary.
Vocabularies and dictionaries may be works in progress, but those works can’t progress far unless thought goes into where the information will be stored so that it will be useful to users. “You need a commitment to funding,” he said, for initial purchase of tools to support metadata and Data Governance, he noted. In addition, decisions about what tools to use also should consider issues such as their support for workflows needed to enable Data Governance.
As an example, Seiner said organizations may come up against instances where a change is requested for something at the technical metadata level – something that may have a cascading impact at the vocabulary level. “We have to be able to involve people in specific workflows in a tool for that, and a lot of the tools now have excellent workflow features that can even include sending email notices to people,” he said. Don’t forget putting policies in place to ensure that best governance practices for metadata and other changes actually are followed. That’s key to ensuring that the right parties actually do get alerted before someone hits up a DBA for a change to a table, before that change’s potential impact up and down the line is clearly thought out, he cautioned.
Clearly, long range planning is important. But, don’t forget about hitting the low hanging fruit mark quickly, too. It’s important to demonstrate value there as soon as possible, he said, “in the way people in the organization will use the metadata and trust the metadata and in how it helps them do their job.” To that end, make sure it’s technically feasible to accomplish these quick-return missions, rather than picking out tasks that are so complex that value demonstration takes too long.
For instance, in most organizations there already are some business terms – benefits descriptions in handbooks for healthcare businesses – that commonly are used by all parties, and likely people at the top of the organization want to see these become formal standards. “Gather a group of people to create a first cut at a vocabulary, get it approved at the highest levels, and that’s a real benefit,” Seiner said.
Finally, don’t neglect how important communications is to the whole culture change of governing metadata. That includes everything from helping end users articulate what they need to use data effectively all the way down to holding stakeholders’ hands as they start to learn to use vocabularies.
“Help them,” he advised. “Change things in request forms to make sure they are directed to the vocabulary so they know what terms to use. Help them gain access to this information.”
