Data Modeling is Data Governance

The name of this article is the name of a Real-World Data Governance webinar that I will be giving in September with DATAVERSITY and special guest Dave Hay. Dave Hay is a data modeler extraordinaire and author of several industry leading books on data modeling. He has contributed many articles to TDAN.com in the past and is very quick to view everything as a data model, a fact that became very obvious to me when we worked together for a high-level government organization several years ago.

When discussing the webinar we both thought the topic would make an interesting article as well. The title should grab the attention of data modeling people as well as data governance people and, perhaps, ruffle both of their feathers enough to gain their interest.

The truth is Data Modeling by itself is not Data Governance – in its entirety. But … Data Modeling is Data Governance. Let me explain.

Data modeling is a data discipline. Through that discipline we design our organization’s data, reduce redundancy, follow standards, and build business-useful definitions for the data. Data modeling actually does much more than that. Ask any data modeler of Mr. Hay, Mr. Hoberman, Ms. Lopez and Mr. Silverston’s stature. They can tell you the value data modeling brings to the organization much better than I can.

Data modeling can be done well, or … less well. Some data models include “cheeseburger” definitions (what is a cheeseburger – a burger with cheese) and some have well thought out and validated business descriptions of data that make that data production and usage infinitely more valuable. The use of data modeling varies widely from organization to organization.

Some organizations have Enterprise Data Models (EDM) that are built to design the entirety of data for the organization. Let me write that again for emphasis – design the entirety of data for the organization. Developing the EDM is often a monstrous task that requires the involvement of a plethora of business and technical people discussing the detailed data and information needs of the organization. Some people view the enterprise model as the place to start the improvement of data and data quality across an organization. Other people view the EDM as a step towards defining and addressing the overall data needs of the enterprise.  Still others view the development of an EDM a big waste of time (no telling for some people’s line of thinking).

Some organizations model data for their internally developed information systems and/or for the data that resides in their data warehouse or business intelligence environment. Often these models are smaller than an EDM and are built for specific purpose – although many organizations select to reuse components of existing models to create new models. Again, data modeling is all about data discipline.

Other organization purchase industry data models, follow described patterns for producing data models, and otherwise take immediate steps to acquire and place discipline around the design phase of defining, producing and using data. Data modeling is, or has been in the past, viewed as the basis of data management activities for the organization.

There are many reasons to create a data model. These reasons include following data standards, reducing redundancy, putting business definition to data, and coming to grips with how to define data better or manage the definition of data as an important asset. There is no doubting that data modeling is both an art and a science but that the primary reason to model data is to instill discipline around defining data for the organization.

Industry definition tells us that data modeling is a process used to define and analyze data requirements needed to support the business processes within the information systems in organizations; while the process of data modeling involves professional data modelers working closely with business stakeholders, as well as potential users of the data and information systems.

According to Steve Hoberman (another data modeler extraordinaire and presently the author of TDAN.com’s The Book Look column, my book’s publisher, and long-time TDAN.com contributor), data modeling is the process of learning about the data, and the data model is the end result of the data modeling process.

So why do I say that Data Modeling is Data Governance?

Data Governance is the execution and enforcement of authority over the management of data. Data modeling can be considered the execution and enforcement of authority over the definition of data. The discipline of data modeling involves the “right” people at the “right” time to define the “right” data for the organization. This is the essence of Data Governance.

Data Stewardship is the formalization of accountability for the management of data. If you subscribe to the idea that everybody is a data steward because of their relationship to the data (a core tenet of the Non-Invasive Data Governance™ approach) then certainly the people providing information and assisting the data modelers must also be data definition stewards. And to think, the people that the data modelers work with have been playing the data steward role way longer then the term “data steward” has been trendy.

Data Modeling is Data Governance – or at least a piece of Data Governance – because it is discipline that is necessary to make certain the design of data is the way it needs to be. Organizations that do not model their data have a more difficult time improving the value they get from their data because their data becomes riddled with inconsistency and misunderstanding. Ask any organization that does not model their data if their data is being governed. The sure answer will be “no”.

That is why I say Data Modeling is Data Governance. Do you agree that Data Modeling is Data Governance? Respond if you do. Or if you don’t. Whether you agree with me or not, or you are just interested in hearing what Dave Hay has to say about my assertion; either way please register for and attend the Real-World Data Governance webinar titled Data Modeling is Data Governance on September 17th at 2pm EST. I hope to see you there.

Share

submit to reddit

About Robert S. Seiner

Robert S. (Bob) Seiner is the publisher of The Data Administration Newsletter (TDAN.com) – and has been since it was introduced in 1997 – providing valuable content for people that work in Information & Data Management and related fields. TDAN.com is known for its timely and relevant articles, columns and features from thought-leaders and practitioners. Seiner and TDAN.com were recognized by DAMA International for significant and demonstrable contributions to Information and Data Resource Management industries. Seiner is the President and Principal of KIK Consulting & Educational Services, a data and information management consultancy that he started in 2002, providing practical and cost-effective solutions in the disciplines of data governance, data stewardship, metadata management and data strategy. Seiner is a recognized industry thought-leader, has consulted with and educated many prominent organizations nationally and globally, and is known for his unique approach to implementing data governance. His book “Non-Invasive Data Governance: The Path of Least Resistance and Greatest Success” was published in late 2014. Seiner speaks often at the industry’s leading conferences and provides a monthly webinar series titled “Real-World Data Governance” with DATAVERSITY.

  • Richord1

    I think this article is a good start but Data Governance remains too abstract as a concept and does not address the underlying limitations of data model designs. Applying data governance to poorly designed data and data models is like trying to improve product quality after the product is in the hands of the customer.

    We need tactical improvements to data modeling. Some of the limitations of data modeling have been documented by David Hays and Michael Brackett. However, data modeling remains a technically dominated practice that has contributed to data silos and databases that fail to meet the business needs. There are two factors that contribute to these results.

    First data modelers and most users are data illiterate. Few understand the meaning of the data and that various context specific interpretations of data; e.g. data semantics and pragmatics.

    Secondly data modelers and users ignore the social and cultural aspects that affect a data model design. The participants involved in the design come with their own values, biases and needs and these are largely ignored in the data model design. For example the sales department has a “bias” on what constitutes a customer; a prospect is a customer. Finance has a “bias” that a customer is an entity that purchased a product or
    service. These are the values that each function embodies. So how does the data
    modeler decide “what is a customer”?

    There are numerous examples of how the values of those involved in the design affect the data model design implicitly. Unfortunately data modelers are not trained or skilled in making these implicit values explicit. This requires a socio-technical design approach.

    There are examples from other industries and disciplines on how to include values in design such as Value Sensitive Design. There are numerous practices and experiences on applying socio-technical design principles to the design of products, services, organizations and manufacturing. Data modeling remains locked in time with the same
    practices being applied for that past 30 years with little improvement in the outcomes.
    This is not a technical problem and data modelers are not trained or equipped
    to address this.

    It is time that data modelers learn from other disciplines to improve their performance. First they and their users must become Data Literate. Second they need to apply
    socio-technical design principles to data modeling. Once data models are improved,
    then and only then can Data Governance become possible and beneficial.

  • Robert Karel

    Hi Bob – A great, provocative article, and I agree with many of your observations. I think an alternate title could have been “Which came first, the data model or the data governance?”

    While data modeling as a discipline is far more mature than data governance, it’s sphere of discovery and influence on the business as a whole remains within a smaller team. Data governance aspires to define and evolve “right” data behaviors across people, processes and systems across the entire information ecosystem – a great vision, but one that most organizations still struggle with. I view the exercises of discovery and definition included in an enterprise data modeling initiative to have strong synergy with the analysis required to build a case for a data governance initiative – so better to join forces than have them as separate, distinct efforts.

    Even in the world of data modeling, those that are looking beyond simply defining entity relationships but truly looking at building a canonical model taking into account semantic alignment and incorporating business and process context are even closer to creating a baseline roadmap that a data governance program could leverage.

    The biggest objections to kicking off data governance with a data modeling exercise would of course include “it’s just IT doing IT stuff, business doesn’t have a role to play” and “why are we taking time defining models instead of solving immediate business problems.” But both objections are easily mitigated with effective program management.

    In my view, experts across both data modeling and DG disciplines can greatly benefit from collaborating with one another. Data governance evangelists can leverage the discovery and business analysis being done by modelers in identifying current state pains and opportunities for improvement, and will have a place to actually document – and enforce – and number of the data management policies and standards they’re driving: In the models themselves. And the data modelers greatly benefit from a DG initiative because when done well, these efforts are building strong business cases evangelizing why business stakeholders must care about data and recommended process changes, business rules and workflows to optimize the best practices that data modelers are identifying.

    Thanks for starting this conversation, looking forward to your webinar!
    Best regards
    Rob

    • Richord1

      The problem remains that the same people trying to “solve” the data management problems are those creating the problems in the first place… IT.
      Few data model designs follow even the most rudimentary data modeling best practices to begin with such as a comprehensive data dictionary and controlled vocabulary. Most designers jump into the physical design before even contemplating anything that resembles a conceptual model. Most modifications to a databases are done at the physical level, ignoring the conceptual and logical models.
      Data Governance is IT driven yet IT does not have the skills to “govern” itself. We don’t see business people evangelizing DG.

      • Robert Karel

        Can’t completely agree with you here.
        First, I think the business is equally if not more to blame for data problems – IT is sometimes an enabler or accomplice, but broken processes are a huge contributor to bad data and the business owns the processes, not IT. IT just aims to automates those processes (on occasion).
        I also disagree with the blanket statement that DG is IT driven and that the business isn’t evangelizing it.
        DG must be a collaboration between business and IT, and the “Driver” can be either – it’s where the most effective program management skills live, and that is often in IT. The important distinction is that the DG Driver is not the owner of the policies, standards and measures of success (think DACI). The business must own the business impact of data, but nothing wrong with IT playing the PM role. And regarding the lack of business evangelism, I agree that failed DG efforts are often due to minimal to no business engagement and ownership, but we’re seeing a ton of successful programs where the business stakeholders aren’t only “evangelizing”, but they are in fact driving the DG investments and priorities.

        • Richord1

          When we begin to associate systemic problems with blame we are on the slippery slope to irreconcilable differences and thus the proverbial gap that people refer to between IT and the business.
          Business processes are the root cause of many problems but what we discover when trying to change business processes is that in many cases they too are systemic problems. Exception handling is the Achilles heel of business processes. IT systems don’t deal well with exceptions but humans do. As a result many business processes require human intervention which IT cannot solve.
          With regards to who is evangelizing Data Governance I think it is primarily the technologists. Business people gave up on evangelizing anything with the word governance after corporate governance proved to be good publicity but not so good for business.
          And with IT’s continued history of failed projects, I suggest they are not the poster child for PM.
          The measure of success of any change program such as data management is sustainability and performance measurement. How many data governance programs have survived based on industry standard metrics? Since data governance has no single definition and certainly no set of industry standard metrics then it would be difficult to determine success objectively.
          In spite of that I believe improved data management is good business but data governance as currently professed is not the solution. At best its window dressing and at worse its a distraction.

          • Robert Karel

            In that, I think we’ll have to agree to disagree.
            Best regards – Rob

          • Richord1

            I don’t disagree, I have a different perspective. My research, observations and hands on experience don’t correlate much with what I read.
            I want proof like I suspect many people want before embracing data governance. Data management is a life time commitment for an organization and should be undertaken with a degree of skepticism.

            And, Mr. Seiner opened the topic to a spirited debate; ” ruffle both of their feathers enough to gain their interest” . Seems like he accomplished that even before his webinar :)

  • KR

    I would tend to disagree with the gist of the article and agree with some of the other comments made here. I think data governance resides outside IT altogether…or it should. When data modeling becomes data governance, it is because data governance did not exist–or was not communicated adequately–before the data was modeled. The exercise of authority over data should come directly from the business side of the house, rather than derive from the modeling process. Sometimes data modeling must fill a void, but I don’t think that’s what we should be going for.

    • Chad Snow

      I think it is helpful to recognize the duality of the full end-to-end data modeling process. Part of the process is a business activity and part of it is a technical activity. When business owners get together and decide what business objects they care about, how those objects relate to one another and some of their key attributes, that is a very business-centric activity and the data modeler must wear their ‘business’ hat in those discussions. This aspect of data modeling is driven by the business and represents a key and early step in developing a data governance framework. Later, when it comes to the physical implementation of those business concepts, the modeler puts on their ‘IT’ hat and optimizes the physical environment for maximum performance.

  • Tim Annis

    Data Modeling is Data Governance in the same way that a Border Collie is a Dog. So your statement is true, but not complete

    • Sharath Purighalla

      He did say “Data Modeling is Data Governance – or at least a piece of Data Governance – because it is discipline that is necessary to make certain the design of data is the way it needs to be.” That’s the same as saying “A Border Collie is a dog, or at least one breed of dogs”. A very true statement. And a sound data model informs the rest of your data governance policy,

Top