If your CIO can’t draw your data architecture on a napkin, you don’t have one.
Here’s a useful exercise for Chief Data Analytics Officers (CDAO), at your next one-on-one with your Chief Data Information Officer (CDIO), hand them a blank piece of paper and ask them to draw your health system’s data architecture.
Don’t beleaguer the usual long discussions on the org chart. The vendor list is not the utmost important either. It’s the architecture – the blueprint that shows where data comes in, where it lives, how it moves, and where insights for decisions are made.
In most health systems, this would look like a city map after earthquake or maze where it’s easy to get lost – boxes and arrows everywhere, sometimes looping back and forth around and leading to multiple data assets.
A separate cluster for the departments, particular specialty areas or health plan. For academia, perhaps another similar one for research. Some even may have multiple Epic or EMR instances that don’t connect. Often landing on a SQL data warehouse (or multiple ones), where only the long timers remember the changes and names. Perhaps so many varieties of legacy and add on technologies, if you put the vendor logos it would look like satellites with no ground station.
Not to say that any technology team or leadership is to blame for this. This is what twenty years of well-intentioned, When you have a problem-by-problem procurement for twenty years with well-intentioned siloed solutions, this is the final result in a complex health system. At those points in time, and problem occurrences – every box was the right answer to a specific question at a specific moment. Nevertheless, the final architecture is not somebody thoughtfully designed.
This makes your job harder (and let’s admit some fun too), as your main focus as CDAO for the next five years would be to replace that map with one that fits on a napkin. Simple and easy to mind map at C-levels. This post is about what should be on the napkin, why, and what conversation a CEO or C-suite leaders should be having to get there.
Why is architecture a C-Suite conversation?
Many would argue bringing the C-Suite including the CEO in an architecture conversation? That’s too much and unnecessary. Traditionally, there has always been a tempting view that data architecture is a technical matter, properly delegated to the CIO and the CDAO, and that the CEO’s or C-suite leaders’ role is to fund it and stay out of the way.
In my opinion, this view is a thing of the past, and expensively so.
Essentially, data architecture determines what the organization can do. It often determines whether a strategic initiative — whether it be value-based care expansion, or a new service line, or a research partnership with a pharma company, perhaps a workforce optimization program — can be executed in months or years. It determines whether the AI investments the senior leadership or board keeps asking about will produce returns or pilots. It determines whether the organization can answer a regulator’s question in a day or a quarter.
If you think about it, data architecture essentially reflects what the organization can do with data. It is an indicator of if a strategic initiative, especially in the analytics or ai driven ones, can be executed in months or years. It would show if an organization can response to a regulatory question within days or needs a quarter.
Now we are at a time, an inflection point with huge potential of data and AI, the systems that move fastest over the next decade will be the ones whose senior leadership understood that architecture is strategy, asked the right questions, and made consolidation a leadership priority rather than an IT project.
It may appear as additional burden on leadership; do they need to transform their style to an unrealistic state? Likely not, as the good news is that the conversation is not technically complex. In fact, without technicality – the modern conceptual model for data architecture fits on a napkin.
Data curation through Bronze, Silver, Gold
Over the last decade data architecture has matured and consolidated towards more unified and simple patterns on modern platforms. Everys healthcare, financial services, retail… every industry that has rethought their data foundation has moved towards a consensus pattern of organizing data into three layers. They may be named differently but the structure is the same. Here are the three layers:
Bronze: first layer is the one for raw ingredients. Essentially, every piece of data and from every source system are ingested ‘as is’ form and stored in this layer without any transformation. Whether it be Epic clinical data, SAP ERP data, Workday HR data. Or Pharmacy systems, imaging metadata, health plan claims. Research source data or IoT device feeds. Everything from everywhere lands here, with original form in full fidelity, with lineage tracked from the second it arrives.
So, why do we need this raw data? The key point of having a Bronze is reproducibility. If there’s ever a question as to “what did the data look like at this moment”, the answer is in Bronze. It also enables building a time series data set. Usually, it is not meant to be pretty, and certainly not designed for any querying or end users. This is more like a system of records upon arrival.
Silver is the semi curated layer, more like prepared meals. This layer is created by conforming Bronze data into clean, modeled, reusable datasets organized by domains like clinical, finance, HR, pharmacy, research, population health and so on. Here the provider master is reconciled. The patient identifier is more consistent across. Here, we begin the enterprise standard principles of modeling – the visit, the encounter, the claim – all in a coherent, consistent, integrated fashion.
The goal of Silver is reusability. Bringing multiple analytics products, various AI models, multiple dashboards and KPIs to draw from the same Silver datasets. This is the only way to produce consistent numbers across the organization.
The ultimate Gold layer: a self-service buffet. This is the final most curated information layer, ready-made data products for business consumption and standardized enterprise KPIs. Here, we have the readmission rate, the length of stay, the revenue per encounter, or risk-adjusted outcome metric for the ACO. The principle is to calculate these once, not freshly every time by every team as they need them. They are calculated once, then governed centrally, and open for consumption from everywhere – thus this is the Gold standard.
The point of Gold is enterprise-wide alignment. When the Sr. Executives – the CFO, the COO, and the CMO walk into a meeting and ask about the same metric; they get the same number. In most organizations, where siloed data assets are around, this is often rare, and the absence of it is the key source of half of all executive frustration around analytics. This is also the final fuel for driving AI. You may be tired of hearing this, but the truth is your AI is as good as your data. Data Quality should be prime in all layers, but the Gold layer makes it the easiest to consume with utmost reliability.
This is the back of the napkin talk. Just three layers, simple visual and narrative – Bronze for fidelity, Silver for reuse, and Gold for alignment.
Why it matters and what it replaces
The reason for this matter is what it replaces. We need to understand what it replaces to really recognize why it matters.
In a typical health system often the data architecture landscape looks pretty crowded; with five to seven domain warehouses – e.g. one for clinical, one for research, one for pop health, another for HR, one for Finance; each often some overlapping duplicate data from another but built independently. They are organically grown so often sitting on different infrastructure, each ending up with its own definitions of basic entities like “encounter” or “length of stay”. Each serving 50 to 100+analytics tools, dashboards, and reports. The same metric likely being calculated in different ways by various departments or groups. Along with these warehouses, there are often hundreds of vendor data feeds, most are point-to-point integrations without much clear inventory or lineage.
In most health systems who are in legacy ‘grown over time’ architecture – the cost of this architecture is not primarily the technology bill, though that is meaningful. The main cost is friction. Every new analytics product takes months because the data has to be re-acquired and re-modeled. On AI front, every AI initiative requires a custom data engineering effort because there is no governed Silver layer to build on. Then again, every regulatory request triggers a fire drill because lineage is tribal knowledge. Every executive conversation about performance includes a debate about whose number is most accurate and timely.
Initially, a Bronze-Silver-Gold architecture does not eliminate this friction in week one. It eliminates the structural reason the friction exists. Over a three-year horizon, the difference compounds dramatically.
To be clear, the Bronze-Silver-Gold architecture would not be like a magic bullet in week one, however it is the starting of elimination of structural reason for the friction above. Over time it will eliminate and begin to lift the rest.
The payer-clinical integration opportunity
If your organization is an integrated delivery system with its own health plan, there is a specific architecture decision worth singling out, because it is the most consistently missed opportunity I see.
In an integrated delivery system, with provider and payer arms, there is a tremendous opportunity for uniquely valuable architecture that may not be on everyone’s radar. The promise of managing cost and quality together across the continuum of an integrated delivery model is often more complex due to perceived regulatory barrier or historical organizational reasons. A unified Bronze-Silver-Gold architecture is an opportunity to fix this disconnect in data. We can get member, provider, and episode entities – all neatly reconciled in Silver layer. Even on value-based care metrics, we can possibly get a Gold layer representing all KPIs used by both clinical and plan sides of the house. The aspiration and strategy of senior leadership gets an opportunity to get real, within the legal boundaries of course.
In nearly every integrated system I’ve assessed, usually the health plan’s data lives in a separate cloud warehouse from the clinical data. There are usually good historical reasons — the plan was acquired, or built later, or sits on a different vendor stack, or has regulatory requirements that argued for separation. The unfortunate result is that the strategic premise of the integrated delivery model — that you can manage cost and quality together across the continuum — is not actually executable in the data, because the data isn’t joined.
A unified Bronze-Silver-Gold architecture is an opportunity – the moment to fix this. Health plan claims become a source feeding Bronze. In this scenario, we can get members, provider, and episode entities reconciled in Silver. Even on Value-based care metrics – we can get it computed once in Gold and consumed by both the clinical and plan sides of the house. The strategy the senior leadership has been pursuing for a decade becomes operationally real.
An Integrated Systems that get this right can unlock a class of analytics and AI that systems with separate payer-clinical data simply cannot do. Our desired architecture gets the quickest bang for value in my view, with this single highest-leverage architecture decision.
Decisions, Decisions: Build, buy, or consolidate
A reasonable question at this point is whether to build the platform internally, buy a packaged solution, or work with an existing footprint.
For most health systems, the honest answer is “consolidate first, then build on cloud-native managed services.”
Building from scratch is rarely the right answer. The components — cloud storage, lakehouse formats, catalog tools, orchestration, governance tooling — are mature and commodity. Building them in-house produces a system you have to maintain forever and a talent dependency you will regret.
Buying a single packaged “healthcare data platform” from a vendor is tempting but usually disappointing. These platforms tend to optimize for one use case (population health, or revenue cycle, or research) and become another silo in disguise. The promise of an integrated solution rarely survives contact with the actual diversity of healthcare data.
The pattern that works for most systems is to pick a cloud foundation — AWS, Azure, or GCP — adopt the managed lakehouse and governance services on top of it and consolidate the existing data estate onto that foundation in a sequenced migration. The advantage of cloud-native managed services is that they are operated by the cloud provider, scale elastically, and avoid the on-prem upgrade treadmill that has consumed so much health system IT capacity historically.
A specific note for integrated delivery systems: if your health plan has already invested in a cloud foundation, treat that as a strategic asset to extend rather than a separate stack to maintain in parallel. The cost of running two clouds is rarely justified by the benefit.
What the C-Suite should be asking
The architecture conversation does not require the C-suite to become data engineers. It requires them to ask a small number of pointed questions and pay close attention to the quality of the answers. Some key ones would be:
First my favorite one: “Can you draw our data architecture on one page?” If the answer takes more than a week or produces more than one page, the architecture is not designed; it is accumulated or think ‘band aid’.
“Where is our Gold layer?” If the answer is “we don’t have one” or “it depends on what you mean,” the organization does not have consistent enterprise KPIs, and you should stop being surprised when meetings produce conflicting numbers.
“What does it take to add a new data source to our platform?” A modern architecture answers this in days or weeks. A legacy architecture answers it in months. The answer tells you how agile the organization actually is.
“How is our health plan data integrated with our clinical data?” If the answer is “they’re in different systems,” you have just identified one of the highest-value strategic projects on the table.
“What is our cloud strategy for data, and how does it relate to our existing on-prem investments?” The presence or absence of a clear answer is a leading indicator of how far along the modernization journey the organization actually is.
In my opinion – these five questions, asked seriously and followed up on, will reshape the conversation between the executive team and the data organization more than any consulting engagement.
The leadership reframe
The reframe I want to leave senior leaders with is this: data architecture is not just a technical asset that supports the strategy. It is the strategy expressed in infrastructure.
We need to remember that fragmented architecture forecloses options. A unified architecture creates them. The systems that will lead in the AI-native era are the ones whose leadership understood this early enough to make the investment while the window was still open.
Honestly, the napkin is not complicated. Bronze for fidelity. Silver for reuse. Gold for alignment. One platform. One source of truth.
A simple test. Ask for the drawing. Look at what comes back. Then decide what kind of organization you want to be in five years. Your strategy and organization success depends on it.
This is my second blog write up in a five-part series on what it takes to move a health system from data fragmentation to AI-native. If you haven’t already, please read my first post — Why Most Health Systems Aren’t Ready for AI — covered the six structural gaps that block enterprise AI and a 30-day honest assessment leaders can run themselves. Coming up next: the Analytics Center of Excellence, and the surprisingly difficult question of what to centralize and what to federate.
As always, I welcome reactions and counterpoints— particularly from CIOs and CDAOs who have lived this migration. I truly believe that the field gets smarter when we trade notes honestly.
I write here in a personal capacity. The perspectives shared are my own and do not represent my employer or affiliated organizations.
Becker’s 3rd Annual Fall Chief Pharmacy Officer Summit is the top tier event that the people leading the future of pharmacy convene. This 3-day event is built specifically for Chief Pharmacy Officers and senior pharmacy leaders. The sessions examine the workforce, financial and clinical priorities shaping pharmacy today. Becker’s is where candid discussions flourish and new ideas thrive. Becker's 3rd Annual Fall Chief Pharmacy Officer Summit
