Why Healthcare Data Governance Fails (And How One Insurer Got It Right)

20 years of centralized governance beats the data mesh hype every time

Imagine you’re the first data engineer who starts at a mid-sized healthcare organization. You’re tasked with modernizing their data infrastructure from brownfield by migrating from on prem into the cloud.

In your first week, you get a lay of the land and realize there are 50 legacy Power BI reports directly connected to the database. The database contains source data mainly from their EHR. There are a few other sources such as call center data and their financial general ledger (probably NetSuite).

You explore and sift through hundreds of stored procedures and agent jobs used to load their data in batch pipelines.

During this process, you uncover Net Revenue SQL logic defined in 3 different ways across two stored procedures and one Power BI DAX formula.

You dig through shared OneDrive folders hoping to find (at a minimum) a stale data dictionary or data glossary - but you find nothing.

Pretty soon you’re on a call with the VP of Revenue Cycle and CEO asking how the company defines Net Revenue.

The VP of Revenue Cycle says deductions include write-offs.

The CEO says deductions do not include write-offs.

A big debate starts on call. You sit there and mediate the discussion while they hash out the “why” for each respective logic. Neither stakeholder is necessarily wrong, but viewing a KPI from a different context can create slight logic differences.

This exact scenario is a result of no central data governance.

This matters GREATLY because it causes strategic misalignment, which erodes company-wide trust in the data. A lack of trust creates poor or no decisions that limits the profitability of healthcare companies every day.

Why Decentralized Governance Doesn’t Work in Healthcare

In 2019 (right when I got into data), data mesh was surfacing as a go-to enterprise architecture for data management.

Snowflake.com defines data mesh as:

In a data mesh, teams actively manage the data within their specific business domains. These teams also build and maintain pipelines that deliver data products to consumers throughout the organization.

Reading that, one can rightfully assume strong governance standards are needed for this approach to work.

News flash: unless you’re a highly data mature organization with lots of investment in your data ecosystem, the data mesh does not work. No mid-market healthcare company has enough investment dollars to throw at the data mesh. And even still, investment dollars don’t guarantee a good outcome.

The regulatory complexity, complex EHR and insurance claim data, and highly sensitive nature of PHI data makes data mesh a failed architecture in healthcare. I can see the positive intentions of data mesh - aka giving control to the users closest to the data. But the reality is, it creates far more complexity than it does value.

Getting a Veteran Senior Data Architect’s Take on Centralization

Last week I had coffee with a fantastic senior/lead data architect at a large insurance company in Columbus, Ohio (my hometown). That conversation inspired the writing of this issue - it was filled with passionate and energetic war stories about our experiences in the space.

To have someone who has been in industry since the early days of data to now, I felt like I was talking to one of the godfathers of data engineering like Bill Inmon or Ralph Kimball. A truly knowledgeable and business-first architect, I left that conversation filled with fire to continue executing for our clients.

His arguments against decentralization were formed through learned experiences. The approach he’s taken at his insurance company is centralization. They’ve experienced a ton of success from that.

The data team owns all of the data. The key partnership is between the data architects to the business stakeholders.

This lead architect is responsible for the data dictionary, glossaries, and all metadata management. He’s spent years building an in-house solution to handle all of this comprehensively.

Any new metric or dimension that needs defined is first documented with comprehensive data lineage right in the code notes. That is then loaded into the data dictionary/metadata explorer he built.

The business definition is then brought before business stakeholders and explained in plain English by the architects. Whatever is agreed on in that meeting is then updated in the data glossary. They’ve been leveraging this process for a long time, and it has worked VERY well.

The glossary can be consumed via a UI. This has resulted in a mature data organization, trusted metrics, and faster decision-making to help their customers.

It’s worth calling out that this insurance company had the budget to build their own solution. Lots of healthcare companies don’t have the time or budget to do that! Buying a solution of the shelf for data governance and metadata is completely fine.

Let’s break the above into 3 pillars:

1. Technical expertise lives where it should (data team understands the data deeply)

2. Strategic partnership with business (data architects bridge the gap)

3. Governance assets that scale (data dictionaries, documented definitions, accessible metadata (built or bought) )

These three pillars work together as a system. The data team’s technical depth ensures governance decisions are sound. The partnership model ensures those decisions actually reflect business needs. And the governance assets ensure the work compounds over time rather than getting lost in someone’s head or a forgotten Confluence page.

Your Next Steps

So what does this mean for you?

If you’re seeing the same symptoms at your organization - multiple definitions for the same metric, stakeholders arguing about what “net revenue” means, data dictionaries that don’t exist or haven’t been touched in years - you have a governance problem. And it’s not going to fix itself.

Here are the red flags that tell you it’s time to centralize governance:

You need centralized governance if:

• Stakeholders are defining metrics themselves without technical oversight

• You have 3+ definitions for the same KPI floating around

• Business users are building their own logic in Excel or Power BI DAX calculated fields

• Nobody can agree on what the “source of truth” is

• Your data team spends more time reconciling conflicting numbers than building new capabilities

Where to start:

1. Audit your current state - Document every place a critical metric is defined (stored procedures, DAX formulas, Python scripts, Excel macros). You need to know the extent of the problem.

2. Centralize ownership to the data team - Make it clear: the data team owns governance. Not the business. Not IT. The data team.

3. Build your first data dictionary - Start with your top 10 most critical metrics. Document the business definition and the technical logic. Get stakeholder alignment. Load it into an accessible tool or UI.

4. Create the partnership model - Pair your data architects with product managers (if budget allows) who can translate between technical and business language. This is your bridge. Otherwise lean on data architects to communicate directly to the business like this insurance company I discussed.

The conversation you need to have with leadership isn’t about control. It’s about clarity.

It’s about ensuring that when the CFO asks “what’s our net revenue?” and the VP of Revenue Cycle asks the same question, they get the same answer. Because right now? They’re not. And that’s costing you more than you realize in bad decisions, eroded trust, and strategic misalignment.

The Columbus architect I spoke with didn’t build a mature data organization overnight. It took 20 years of discipline, iteration, and commitment to centralizing governance where it belongs - with the people who understand the data on a technical level.

But you don’t need 20 years. You just need to start.

---

Christian Steinert is the founder of Steinert Analytics, helping healthcare & roofing organizations turn data into actionable insights. Subscribe to Rooftop Insights for weekly perspectives on analytics and business intelligence in these industries.

Feel free to book a call with us here or reach out to Christian on LinkedIn.

Also - check out our free Healthcare Analytics Playbook email course here.