After a failed first attempt at creating a data warehouse, the University of Notre Dame, Ind., built a data governance structure designed to empower members of the campus community to use analytics to make data-informed decisions.
In 2013, the University of Notre Dame (ND)—a global, private research university inspired by its Catholic character and with more than 12,500 students—deployed an enterprise data warehouse as part of a data-informed decision-making initiative. “After several attempts of mapping staff-related data into the data warehouse, we found that what was mapped would meet human resources’ or institutional research’s needs, but would never meet both,” explains August “Augie” Freda Jr., campus data steward. “The two offices struggled to come to agreement, and the fledgling initiative stalled.” What followed was a tough and candid self-assessment that prompted the institution to bring in a consulting firm that offered this advice: Make data governance the foundation of all analytics endeavors.
So, in 2015 Notre Dame launched a new analytics initiative. Mike Chapple, who was then senior director for IT service delivery, led the way. Chapple, who is now the academic director of the master of science in business analytics program and the associate teaching professor of information technology, analytics, and operations, recognized that a successful data governance process had to be top-down-driven and grassroots-executed.
“Under Chapple’s guidance, ND tried to be as inclusive as possible by letting the different units on campus decide if they wanted to be the stakeholders who would help define the metadata of our institutional information,” Freda recalls. Initially, 17 people, each representing their respective functional units, were identified as stakeholders to participate in creating this metadata.
Many of the stakeholders, new to the process, initially wondered, “What’s all this about, and why is it necessary?” Some commented, “I spend my day in the transactional system, and I get what I need. This doesn’t apply to me.” Others voiced a common refrain: “It’s my data. I captured and created it. I have concerns about sharing it.”
“We quickly realized that we needed to develop a culture in which data stewards—and other campus leaders—understood that neither individuals nor individual units own data, rather that data belong to the institution,” says Shannon Cullinan, executive vice president. “To support this philosophy, we created a data governance structure with formal and informal components.”
The formal component consists of the following two committees:
- The information governance committee is charged with information security, data classification, safe handling of data, and data governance. A standing university committee, its members include a handful of senior leaders, such as audit and general counsel, and data stewards from each domain of data, such as HR, student academics, student life, and so forth.
- The dataND steering committee includes the executive vice president, vice president of finance, provost, and a number of other vice-president-level leaders, as well as many members of the information governance committee.
“The steering committee provides guidance and breaks down barriers for our implementation of the data warehouse and for the use of institutional data for making better decisions, and prioritizes where to allocate resources,” Freda says. “At some point, when our data-informed decision-making program is self-sustaining and part of our institutional culture, we plan to declare victory and disband this committee.”
The informal component—the data stakeholders—now consists of 30 to 35 people from across campus who handle the day-to-day governance activities, such as the creation of all the metadata, terms, and access rules. The members of this boots-on-the-ground group, which includes data stewards and key data users, can self-select in or out of meetings, depending upon one’s stake in whatever the subject of discussion.
“This lets us engage the stakeholders in a meaningful way, while also recognizing their other obligations,” Freda says. “We work to make sure using data isn’t an add-on to other campus responsibilities, but that data are seen as a tool to propel everyone’s efforts.”
Guiding Data Stewards
Creating the data governance structure wasn’t as tough as getting people to agree to the new philosophy of sharing data. “That was really our biggest challenge from Day One,” Freda emphasizes. “Everybody wanted data but hesitated to share the data we required to build a warehouse and to empower members of the campus community to use analytics to make data-informed decisions.”
He cites this example: In October 2015, ND was ready to go live to early adopters with 500 terms and definitions representing student, course, faculty, and staff data when he discovered that the stewards had designated 70 percent of the data as “restricted access.” This meant the data couldn’t be shared widely.
Going directly to the top, Freda forewarned John Affleck-Graves, who was then the executive vice president, that the program was ready to launch, but users would not see anything of much value because 70 percent of the data had been restricted.
“Why? What kind of data are we talking about?” Affleck-Graves asked.
“Upcoming courses and their instructors are examples of the restricted data,” Freda replied.
“Why would anyone restrict such routine information?” Affleck-Graves asked.
Repeating what the data stewards had told him, Freda responded with answers ranging from “The data could change” to “The data could be misinterpreted” to “A piece of the data is missing.”
“None of those are good reasons to restrict data,” Affleck-Graves said.
What appeared at that moment to be an insurmountable obstacle became a turning point, according to Freda. “In our efforts to build our data warehouse and data governance structure, we had failed to provide our data stewards with concrete guidelines for releasing data,” he explains. “Because they didn’t recognize there was an acceptable level of risk in having information more broadly available, they retreated to the safest decision: restriction.”
As a result, the executive vice president, general counsel, senior associate provost (representing the provost), and vice president of strategic planning and institutional research (representing the president) met with the data stewards to give them guidance on how to make future decisions. “We expected it to last 15 to 20 minutes,” Freda remembers. “An hour and a half later, we came out with Information/Data Access and Security Guiding Principles, a two-part document that we still use.” (See sidebar, “Setting Guidelines.”)
Freda points out that the group of leaders at this meeting reconfirmed the role and authority of data stewards and emphasized that nobody was challenging their authority or responsibility. “As a result of the meeting, we flipped the burden of proof from the party wishing to gain data access to the party wishing to restrict access,” he says. “We decided to assume the data would be accessible unless there is a good reason why it shouldn’t be. When we took that back to the data stewards, the percentage of restricted data dropped from 70 percent to 20 percent.”
This decision represented a real cultural signal, Cullinan says, not only to the data stewards but also to all faculty and staff, that data are an asset of the university and not of a particular unit, and that ND is committed to opening access. “It empowered the stewards to know they could share access—or to know they would be pushed to share access,” he says.
Trusted Takeaways
The objective stated in Notre Dame’s Information/Data Access and Security Guiding Principles is: “To deliver the right information, to the right people, at the right time, to support the right decision.” On the pathway to this data-informed decision making, Notre Dame’s faculty and staff faced challenges specifically when it came to transforming the culture around data, resulting in valuable lessons that have continued to inform the data-governance process.
“As change agents, we have learned to be patient but firm and to create the opportunity for many perspectives to be voiced in order to build empathy within the campus community,” Freda says. “For example, at workshops, I resist the urge to have the right answer at the start. Instead, I let the vesting parties define and decide terms, which encourages future participation and garners buy-in [from] across campus by giving people value.”
Other lessons learned:
Don’t skip step No. 1. “For us, data governance was a means to an end,” Cullinan explains. “The first time we tried to deploy any data or tools in the business intelligence environment, we couldn’t get agreement. If we had three people in the room, we had three completely different opinions and needs for using a particular data element. We learned from that first failed attempt that if we didn’t deal with the data governance issues, we would never reach our goal—data-informed decision making.”
Find a senior-level champion. “While we built a broad section of stakeholders who had bottom-up support, we also had the necessary champion in our executive vice president,” Cullinan continues. “John [Affleck-Graves] only had to say, ‘This is a priority. Data-informed decision making is where we are going and is important for our institution,’ for people at all levels to pay attention. Because John became a champion, if there was a meeting once a month, you pretty much didn’t miss it. With a senior leader at the helm, the campus understood the value of investing time in analytics.”
Expect a certain amount of drama. When ND decided to examine headcount growth, which can be a sensitive subject, Vice President of Strategic Planning David Bailey coined the expression “Drama doesn’t equal damage.” By exposing the statistics around headcounts and allowing the subsequent drama to manifest itself, the institution ended up with significantly better, more accurate data, according to Freda. “In fact, we found several cases of individuals associated with the wrong units from a budget standpoint,” he says.
Value people’s time. Notre Dame allows stakeholders to opt in or out of meetings to hash out terms. Freda sends out a poll saying, “We’re going to talk about these 15 terms,” providing rough-draft data definitions, so they have context. Stakeholders reply: “Yes, I need to participate,” or “No, I don’t need to participate; just inform me of the outcome,” or “No, I have no stake in this.” This approach helps us keep from wasting people’s time, notes Freda. “Plus, we’re not wrestling calendars—it’s much easier to schedule four people for a meeting than it is 35.
A couple of times, Freda has convened a meeting and realized a key stakeholder is missing. In those cases, the discussion of those terms is put on hold until the right people are in the room. “This transparent procedure ensures that everybody knows what it is we are talking about and can contribute,” he says.
Define limits. Early on, Freda tried to focus on use cases. He would ask, “What questions are you trying to answer?” “The problem with that,” he says, “is every question became a project. We decided we really wanted to flip this so we were in the position to answer the questions that we didn’t know to ask yet.”
For example, when ND started to build out data about facilities, rooms, and spaces, the campus safety representatives became excited, saying, “Wait a minute. Do you mean that we can use course registrations and Google Meet to tell us where individuals are located at certain periods of time?” Now they know that on Mondays, Wednesdays, and Fridays at 9 a.m., these particular students are scheduled to be in this room in this building. In the case of an incident response, that’s critical information.
“To get the answers to unknown questions, we have to have consistent data and be confident in what and how we are counting,” Freda says. “The governance component becomes the foundation of institutional strategy.”
Data Costs
Notre Dame’s data warehouse now contains more than 1,000 data definitions pertaining to students, courses, staff, faculty, physical space, and endowment, and the institution is in the process of moving in budget and budget-management data.
“Although we still struggle with access issues, it’s clear our culture is moving to more accessibility,” Freda says, adding that some wins along the way have been helpful in turning potential naysayers around.
Cullinan, who was the vice president for finance before becoming the EVP last July, explains that he is occasionally asked whether the cost of the venture into data-informed decision making is a good return on investment. “I firmly believe that to address the cost curve, which we all know is one of the bigger challenges in higher education, we need good analytics. For us, the upfront and ongoing costs of the business intelligence team, as well as leadership’s investment, were always a bit of a false choice. We need accurate data from across the university to address the cost curve long term.
“The more apt question might be, ‘What does it cost not to do this?’”
MARGO VANOVER PORTER, Locust Grove, Va., covers higher education business issues for Business Officer.