How Organic Valley Established Business-driven Data Governance (And Why it Matters in Healthcare)

My Folder

“Kyle and I worked together on the EDW team at Allina Health before he joined Organic Valley as a data architect. Over the past year or so, he has been instrumental in developing a successful data governance program for his new company. In one of our conversations, I shared that healthcare continues to struggle with data governance, and would probably benefit from hearing what has worked in other industries. Kyle agreed to share the Organic Valley story on how they struggled with weak data governance, how they solved the problem, how the solution relates to the ways in which healthcare is currently struggling with data governance, and how healthcare organizations can get started on a similar path to what Kyle did.”

– Mike Doyle, Vice President, Business Development, Health Catalyst

data governance dairy farmWhat do dairy farming and healthcare have in common? Both industries can reap significant benefits from using data to improve performance. I spent a large part of my career in healthcare, helping develop an enterprise data warehouse (EDW) and data-driven processes to improve the quality and cost of care at a large integrated delivery network in the Midwest. Recently I transitioned to a new industry, taking a position on the fledgling EDW team at Organic Valley, a farmer-owned cooperative with just over 800 employees.

The problems we faced at Organic Valley will sound familiar to anyone involved in healthcare analytics. We simply didn’t have enterprise-wide, reliable data that executives could trust. Instead, enterprise-level decision-making relied on department-specific reports that often conflicted with each other. Even the most basic questions—such as Who are our customers? What groupings do we use to report on and analyze our product profitability?—had various answers depending on which area of the business was reporting. Our analytics were further hampered by proliferation of an uncommon language, inconsistent business rules, and various sources of information for the same data entities. Because of this data confusion and the resulting distrust in the data, leaders would end up making decisions based on gut instinct rather than business intelligence.

As we devised our analytics strategy at Organic Valley, one of our foundational priorities was to establish a data governance program. Along the way, we learned several lessons as we established this program—lessons that can apply to data governance in any industry.

Well-intentioned but Ineffective Data Governance

When I arrived at Organic Valley, the co-op had already gotten the ball rolling by forming a data governance committee. Two problems with this committee became clear very quickly:

  • First, there were simply too many cooks in the kitchen. More and more people kept getting invited into the committee until we had over 60 members. This large membership made for a costly data governance program in terms of employee time. More importantly, the involvement of so many people in discussing every change greatly increased the time required to make data governance decisions.
  • Second, the committee was predominantly populated by IT personnel rather than business decision-makers. Technical details surrounding governance decisions were hindering a business-driven data governance process.

As chair of this data governance body, and as someone with experience hammering out data governance in the healthcare environment, my job was to take the lead on clearing these problems.

Creating New Structures and Processes

Our first task was to create a new data governance program structure to replace the ineffective data governance body. The new structure was threefold:

  1. Executive sponsors. Our initiative required support and leadership at the executive level. Because they were most affected by the burden of not having good data governance, we chose the COO and CFO to act in the sponsor roles. Both are data-savvy individuals, which also made them ideal sponsors.
  2. A data governance committee. This committee is responsible for the designation of policies and processes. They are also in charge of designating people to act as lead data stewards (those who will execute those policies and processes). The committee consists of nine leaders from various parts of the business and three IT representatives. Note that the committee has three times as many business leaders as IT personnel. That is a critical point and one that I will talk about in more detail further on.
  3. A data harmonization committee. This group is responsible for executing the policies and procedures of the data governance committee. They perform the day-to-day work of data governance. The committee consists of 16 lead data stewards—two for each of the eight major data subject areas for the enterprise—and two data architects. Each pair of lead data stewards works with the data architects to develop the enterprise conceptual data model.

Thoughtful selection of lead data stewards was very important. Several of the leads had already been acting as stewards of much of the information within their subject area. To get them formally started in the role, we provided additional training. Most importantly, we gave them a data profiling and quality tool to empower them in their work.

Wading into the Water rather than Diving in

We allowed our data stewards to take the process one step at a time. For example, the initial pressing task for the data harmonization committee was to define our data terms. Because we didn’t want to overwhelm the stewards, we chose a definition target that represented an immediate need while designing and developing the data warehouse (e.g., how do we define customer, farm, & product). Next, the lead data stewards were engaged in the design of the conceptual and logical data models and provided consultation to assure that our enterprise-level business rules were reflected in the design of the data warehouse.

As the data governance program matured and the cooperative’s awareness of the lead data steward role expanded, more and more groups began to leverage the governance resources. Now, rather than having an overpopulated data governance meeting every two weeks to provide ineffective consultation to a variety of project needs, there is an empowered set (or subset) of stewards that can instead be invited to the various project discussions as consultants.

Over time, lead data stewards’ responsibilities have matured as they have been engaged by various project teams for specific data tasks. These tasks include such things as assigning security classification to data (with a particular emphasis on our most sensitive data as mandated by the data governance committee) and consulting on data inception or change decisions. One year into the project, we’ve reached the point where these lead data stewards are being recognized across the co-op as valuable consultants.

Our lead data stewards are just beginning to develop their own cabinet of data stewards related to the data subject areas they represent. This additional layer of stewards enables the leads to delegate stewardship responsibilities as necessary. It also covers knowledge gaps and opens a two-way line of communication across the departmental spectrum.

Making Data Governance a Business—Rather than IT—Concern

Good data governance at Organic Valley has required not just a procedural but a cultural shift (and I believe this is true of any organization). We effectively had to change everyone’s mindset about data. The mantra that our co-op needed to understand and embrace was: Data is about business—not IT.

Effecting such cultural change didn’t occur overnight. It took work, planning, and executive support. In addition to creating and supporting our new governance structures and processes, we carried out an internal PR campaign to rebrand data governance. We encountered some aversion to technology intrusions to business that had been running well over the years—but this was outweighed by the frustration of not getting consistent answers. So we tried to address that pain point and to promote the cultural change in fun and instructive ways. For example, we created a video about a Data Superhero to highlight the power of good data governance.

Giving ownership for the quality of data to business people was a cultural shock for them. Some business leaders were already trying to use data, so they were thrilled to have a hand in the process. The majority were more hesitant. However, over the past year, as we’ve gone through the process, business leaders have embraced the idea of making the upfront decisions about data governance (in fact, the data governance program has gained a lot of credibility in the co-op because it puts decisions in business leaders’ hands). IT’s role in the process is to consult with business leaders on the governance decision, to say what is possible and talk about the practical details of implementing that decision.

I cannot emphasize enough what a difference business ownership of data governance has made so far for us at Organic Valley. In fact, though our EDW itself is still in its infancy, establishing this foundation of business-driven data governance has prompted greater recognition of the importance of data and its role in determining the future of the cooperative. Business leaders are no longer disengaged from the data, and therefore they trust it and value it more. Such a foundation sets us up very well for the future as our business continues to grow.

About the Author:

Kyle Bartelt has seven years of experience as a data architect. In his current role at Organic Valley, in addition to data warehouse architecture, he co-chairs the Data Governance Committee and coordinates with the Data Steward (Harmonization) council. Currently he is helping to lay the foundation of a robust data governance model focused on the rigorous collection and approval of data definitions and policies. Prior to Organic Valley he worked as a data architect in healthcare data warehousing for one of the largest healthcare systems in the Midwest. Kyle has an M.S. degree in Soil Science from the University of Wisconsin – Madison and a B.A. degree in Mathematics from Luther College.

Loading next article...