The authors highlight common challenges surrounding data management and governance in today's higher education institutions by offering both tongue-in-cheek and serious best practices.
We've all been there. It is 90 minutes into a 2-hour meeting focused on data definitions, and the 10-plus representatives in the room still cannot agree on what a single business term means for the IT organization. A timeout is called. A participant boldly asks, "Why are we really here?" The response is simple—yet infinitely complex. A campus leader went to a session at a conference and has come back with a goal to bring "data governance" to the organization.
Clearly, this approach to data governance is not ideal. Below we offer 10 other not-so-best practices. Our goal is not to focus on pain points but, rather, to lay out common areas that need to be addressed. If your IT organization is truly serious about implementing a robust and healthy data-governance environment, be sure you don't mimic the strategies below. Doing so will keep your organization on a path toward data confusion and unclear data management.
Top 10 Not-So-Best Practices
#1: Work hard to keep data-governance principles secret.
Do your part within your organization to maintain a confused state around data governance. Proceed with caution when helping others to elevate data literacy. Ensure that data definitions, metadata (data about the data), and architectural solutions to ensure data-management outcomes are either underground or nonexistent. The best practice is to operate in a silo and refuse to speak openly with colleagues about how data that is commonly utilized across the operation. Additionally, given staff turnover rates, don't worry about conducting exit interviews with subject matter experts (SMEs) to document important definitions.
#2: Be sure to immediately launch a data-governance committee.
When most IT organizations think about data governance, the knee-jerk response is to convene a committee of folks who already have more on their plates than they can handle. We believe this is a fantastic idea as a way to sink a data-governance effort. We also believe that forcing these individuals into a room at least once a month for several hours at a time always works. Even better, be sure to choose individuals who know only about the functional nature of the data but not where it comes from, where it resides, or anything about its quality.
As far as agendas go, we would also advise you to pick the most contentious data first for discussion, since building consensus is always easier when committee members spend several weeks trying to define one term. Regarding the committee's work, under no circumstances should you provide a tool for the capture of data-governance work by the committee (more on that later, see #3). Finally, be sure not to provide recognition, incentives (meals), or any other means of positive feedback to the individuals tasked with this work.
#3: Do not fund technology to support this work; basic spreadsheets are good enough.
We do not advocate for finding new sources of funding to generate a high ROI around the data. Instead, a sound approach is to invest in shiny new tools for visualizing data, to spend money on statistical software, or to fund approaches that are duplicative of functions already within your organization. Clearly, a spreadsheet will do the trick (especially if you want to be sure that no derived value from the exercise is achieved in the organization).
Data governance works best when done in isolation. The technology to catalyze these efforts should be equally isolated. Never use a tool that others have access to; this will serve only to help educate others in the organization. Finally, do not advocate for these efforts with a dedicated budget. This practice will allow you (1) to feign ignorance when asked about funding and (2) to proclaim that there aren't any funds available for use.
#4: Ignore key staff in the technical realm and focus on the functional SMEs.
Data governance is guaranteed to succeed when large groups of stakeholders are not included. Because most IT organization conference rooms hold only 10–14 people, you should intentionally stack these committees with functional SMEs. And since data appears via magic, as long as it gets to the appropriate unit, that should be all well and good. By cutting out IT professionals from your data-governance process, you effectively guarantee that you won't be bothered by the technical systems, processes, or architecture. Keeping IT practitioners in their offices is also a sure-fire way to ensure that this expertise is not applied to the benefit of the program.
In alignment with this belief, one way to immediately ensure that a data-governance effort will succeed is to intentionally not engage your chief information security officer, chief information officer, and chief technology officer. These leaders have enough on their plates, and their contributions would only work against you, given that they would build trust in the governance process. Finally, a bonus tip is to ensure that IT SMEs are not involved in any vendor conversations: their roles work best when they are asked to implement a solution after a decision has been made.
#5: Be sure that an executive stakeholder is never involved.
Killing a data-governance effort may be easier than you think. Executive stakeholders have the uncanny ability to align human, financial, and technical resources with the mission of their organizations. This usually leads to disruptive change for you and your colleagues. Executive sponsors also have direct access to large-scale communication outlets, vehicles often leading to coordinated messaging that can be risky if you want to maintain the status quo.
We believe data-governance efforts work best when led by just a few members of the IT organization (or even by a single champion) who care mostly about the data they use and not necessarily about the totality of the data being used across an organization. Finally, if for some unfortunate reason you are tasked by an executive sponsor to do this work, try your best to provide only sporadic updates as a way to help ensure that this latest flash-in-the-pan idea passes quickly.
#6: Ignore any and all assessment approaches to track your data-governance progress.
Under no circumstances should you tolerate either direct or indirect assessment of the governance process. Working in an environment in which you won't be held accountable for your data-governance efforts helps ensure that any progress made is modest, at best. Additionally, assessment requires data-governance efforts to set strategic goals and outcomes. We find that data governance works best in an unstructured, nonstrategic, and scatter-shot format. Likewise, never broadcast information about the data-governance efforts and or about gains made in these efforts as this will undoubtedly backfire and cause others to want to govern even more data. Finally, if you are cornered into purchasing a solution to assist with your data-governance work, make sure that the tool cannot export data about the governance process. Seek out solutions that are repositories only and that are disconnected from the analytic, customer relation management (CRM), finance, or human resources systems in your institution.
#7: Never provide for additional staffing to lead data-governance efforts.
If something has worked for several decades, why change it? A truly ingenious way to ensure that a new data-governance effort will struggle is to give additional responsibilities to current staff in your organization. Providing more resources in the terms of a dedicated professional who is tasked with data governance only paints your organization as one that takes the work seriously. And why would any IT organization want to innovate at a time when everyone is satisfied with the way things are?
Most importantly, by not having a dedicated leader for this work in your organization, you have the opportunity to continue to blame any problems on the current demands and pressures of the daily business needs. It can be a slippery slope when IT organizations contemplate adding dedicated data-governance staff. Carving out a new role takes a conscience decision to disrupt how people interact with each other. Disrupting organizational culture can lead to many other unintended changes, including collaboration, trust, and transparency. For those of us who are happy with our current data landscape, these are dangerous changes.
#8:Do not spend your time creating data-governance communication outlets.
Another easy way to doom a data-governance effort is to allow users across your IT organization to continue to be sideswiped by changes that happen in the data. Data literacy is over-rated. As your organization's data evolves over time, you should let the chips fall where they may regardless of the severity of the outcome. Having a formal process to elevate changes in an organization's data and creating creative and consistent communication outlets will only encourage members of your organization to make more data-informed decisions. These decisions will disrupt and poke holes in existing practices based on anecdotal evidence. Finally, if your leadership forces you to communicate data-governance outputs, make sure that you do so inconsistently and without creativity, as a way to maintain confusion while appearing to be transparent.
#9: Prioritize routine and hand-coded efforts over automation opportunities for data governance.
Ignoring innovation in the data-governance and analytics field takes dedicated focus and a willingness to ensure you spend time focusing on routine tasks. Data governance will thrive when every member of the data-governance process has to keep manual, isolated, and disconnected tools. This practice will ensure that these SMEs track and keep their data-governance work in the silos in which we all currently thrive.
Likewise, expecting data professionals to have to learn a new IT data-governance tool pushes the boundaries of what an organization should expect of an employee's professional development. Ensuring that your technical and functional SMEs continue to stay static in their skillsets goes a long way toward maintaining the IT organization's current level of data maturity. Finally, keeping your data-governance process nonautomated allows SMEs to ensure that these tasks can be pushed to the side to focus on more important tasks, without any fear or repercussions from leadership.
#10: Always launch analytics products without linking them to data-governance processes.
Dazzling your organization's users with beautiful visuals is a great way to distract them from the underlying data. Spending considerable time developing an analytics product and ignoring data-governance processes will help you get these important deliverables to market faster. Most leaders, users, and consumers of data are comfortable with their organization's data and typically will never question where it comes from, whether it is valid, or what it means.
In today's fast-paced environment where we need data quickly, in a way that resonates, data-governance practices should not impede the delivery of these insights. We should trust our users to understand what we are presenting, and we should stand boldly against requests for explanations of the data and how it was derived. If you are currently using data-governance tools, be sure to keep these tools disconnected from your analytics products and be especially wary of tools that allow users to simply click on a data element and find out more information.
In Conclusion (?)
We know that following our guidance above will take effort, but IT organizations that do so will have a very easy time of ensuring that data-governance practices never take hold. This advice can help drive mediocrity and an ever-present culture of ignorance around data validity, data integrity, and data sourcing.
Casting sarcasm aside, and in the spirit of truly helping IT organizations mature in this work, below we offer 10 "real" strategies for success. At the University of North Texas (UNT), three of our previous data-warehousing efforts struggled. In 2015, staff were tasked with implementing a new approach to data warehousing and analytics. On further review of the past, we identified data governance as a common contributor to the previous struggles in this area, and we redefined opportunities for institutional research (IR) to contribute. In early 2018, we had launched our Insights Program, offering training across campus. UNT now has a foundational set of practices to help instill trust and data literacy within our analytics program.1 To achieve success in the Insights Program, we leveraged the following top 10 best practices.
Top 10 Best Practices
#1: Find leverage points and stick to them.
After UNT purchased shiny new business intelligence tools, we thought: "Now what?!" At that point we had to grapple with how we were going to roll out the new tools and make them stick within our university community. We wanted our new data-warehousing and analytics effort to be different from previous implementations, so we committed to two important premises that continue to help us tremendously. First, no analytic dashboard will be released without being fully documented and governed. UNT capitalized on institutional demand for the new analytics environment and users' desire to see their data in the new system as a way to build relationships and work on data governance and documentation. Second, institutional data is exposed as it exists in the source system. Doing so has highlighted inconsistencies (e.g., changing college names, duplicate faculty names, multiple sources of information) and a growing interest in data quality.
#2: Identify a group of champions and grow them.
First off, find your executive sponsor. Initiating and developing a data-warehousing, governance, and analytics program is much easier to do when you have an executive sponsor at a very high level in the institution. Identifying a champion who believes in your efforts is critical. Because of our president's engagement, UNT was able to more efficiently deliver accurate, timely, and dynamic data resources with documentation into the hands of decision-makers. The executive sponsor and also other champions should know the history and the goals of the program and should be willing to share their excitement about the program and talk about its successes. Once your group is established, do not rest: continue to grow your champions, not only in number but also in terms of access to data domains.
#3: Celebrate progress at every opportunity.
Use every chance you have to celebrate program successes. From releasing your first dashboard, to training your first user, to surpassing milestones in dashboard production and the amount of data that is documented, be sure to applaud successes and your champions who helped you get there. The tools and technology, by themselves, won't help you succeed. Focus on people, and highlight those who have joined in the process to evolve your institution's data landscape.
#4: Become a data analytics and governance evangelist.
Believe and communicate that governance and analytics are inseparable. Why go through the effort to build fancy analytics if there is no common understanding about what should be shown? The beauty of self-service analytics is that users can access dynamic information to meet their needs without your interaction. On the other hand, the kiss of death for any analytic tool occurs when it is understood differently by multiple users. Focus your conversations around the need for data governance by connecting back to data reliability and consistency. Do not be afraid to share the truth as clearly as you can. This is the time for unabashed clarity around your work and the impact it will have on the people you serve.
#5: Don't underestimate branding and the power of a great visual.
Your data-warehousing governance and analytics program needs to have a consistent and attractive face. UNT developed simple visuals that helped explain complex concepts to tell our story. As the old adage goes, a picture tells a thousand words! Program diagrams and visuals helped us explain our program to key stakeholders and users, including the key premises, the importance of data governance and documentation, and the overall goals and direction of the program. Branding is critical too. Be sure to have a consistent look and feel for all aspects of your program, including communication in all forms and training announcements and documents, and use your branding on the analytic tools themselves. Users should immediately know when an email, document, or graphic is associated with your program.
#6: Implement, and become an expert in, the data-governance tool of your choice.
Tools matter. Governance requires financial support, and great strides can be made in data governance by focusing institutional resources across functional area, starting small, and providing training and support in a data-governance tool. UNT succeeded early on by first identifying the technical and functional subject matter experts (SMEs) and the data steward for each core area of data and providing them with access to our tool. The beauty of having a data-governance tool is that defining data resources is no longer the work of a data-governance committee, whose members are required to sit in a room for hours in order to define individual terms. Rather, SMEs and data stewards can contribute to an online repository independently—on their own time. Giving them access to the systems and resources they need while they contribute technical and functional documentation was a powerful contributor to the program's success. UNT's use of our vendor's abilities in this area contributed to ensuring that data governance was a foundational first step for the rest of the program.
#7: Develop a way to track progress and set goals for your data-governance program.
Data-governance tools are typically great repositories for term documentation and metadata, but the built-in reporting functionality usually leaves a lot to be desired. You should develop a reporting method that allows you to output and monitor term review information. We suggest further incorporating your data governance and analytics by producing a business term dashboard to track progress and set goals related to business term review. At UNT, this has helped us communicate with administrators about outstanding documentation and needed term revisions and has allowed us to track and celebrate progress as we continue to document and define a greater portion of our institutional data resources.
#8: Piggyback data governance on data modeling and analytics efforts.
In building a data-warehousing, governance, and analytics program, UNT found that members of our data-modeling and analytics teams had to work across the institution as they developed accurate, validated data models to feed analytic tools. In order to meet the needs of our users, UNT partnered with SMEs and data stewards to understand data resources and build key data structures. Forming relationships with our key data-governance representatives provided the perfect opportunity to grow capacity in data governance. UNT could now demonstrate how our analytic tools and their integration with our data-governance program best served the institution. We were able to provide the documentation for all data elements used in data models and analytic deliverables. We easily and readily made these resources available for our 500-plus users. UNT intentionally took every opportunity possible to work with key technical and functional SMEs to maximize our overall capacity for thorough documentation and trusted governance protocols.
#9: Use data governance as a vehicle to eliminate institutional silos.
To become successful, UNT intentionally utilized the governance and documentation process to break down data silos within our institution. Data governance, by definition, forces an institution to identify the data stewards and key SMEs responsible for the technical and functional implementation of data structures. We believe that our efforts to promote the idea that data governance needs to take place within small, functionally based groups were a key to our success. However we also developed, in parallel, structures to resolve differences in the definition, understanding, and usage of key business terms. This process created intentional conversations that clarified long-standing business confusion. By refusing to do data as normal, UNT was able to leverage the strengths of our SMEs without the consequence of further isolating our existing silos of excellence.
#10: Be firm with commitment to data governance but flexible with roll-out and timelines.
At UNT, our first priority as a higher education institution is to recruit, enroll, retain, and graduate students. Through our team's work, our staff relied on the UNT mission, which calls upon us to provide education to change the lives of the students and families we serve. We leveraged this belief to recontextualize the importance of, and our commitment to, data governance. That being said, there are times when functional areas simply cannot invest the bandwidth to prioritize data-governance work. UNT bridged this gap by adhering closely to the commitment to data governance and documentation and, by extension, continuously championing and supporting data-governance efforts under an overall philosophy of demonstrating flexibility with specific timelines and target dates.
In Conclusion (Truly)
By leveraging these principles and practices (and avoiding those we know will harm efforts), UNT has broken through many of the common data-governance barriers facing institutions of higher education. As a result, UNT now has clarity around key business data and the value it affords. The institution is far from being done in this regard, but every new business term or data process under governance results in a positive step forward for our students, faculty, and staff. Intentionally refocusing data governance away from purely an IT or IR initiative and emphasizing its role in helping all decision makers improve outcomes has worked well. Taking our second set of advice and guidance into counsel could enable this level of transformation at other colleges and universities also. Doing so requires personal commitment and a willingness to change our institutions for the better.
The University of North Texas System, IT Shared Services, received a 2019 CIO 100 Award for innovation in technology [https://www.cio.com/article/3391918/2019-cio-100-winners-celebrating-it-innovation-and-leadership.html]. This distinction was awarded based on collaboration with the UNT Insights Program IR staff. In addition to other success factors, data-governance innovation was a major driving force in this recognition.
Note
- See "UNT Insights Program Makes Student Data Easy to Understand and Access" [https://financeadmin.unt.edu/unt-insights-program-makes-student-data-easy-understand-and-access], UNT Division of Finance and Administration (website), accessed August 2, 2019. ↩
Jason Simon is Associate Vice President for Data, Analytics, and Institutional Research at the University of North Texas.
Dan Hubbard is Director of Data Management at the University of North Texas.
© 2019 Jason Simon and Dan Hubbard. The text of this article is licensed under the Creative Commons BY-NC-SA 4.0 International License.