Architecture · Data warehouse vs. operating record
A warehouse stores what every system said. An operating record decides which number counts.
A data warehouse stores copies of data from many systems so they can be reported on and analyzed. An operator-controlled operating record decides which value governs a decision: it keeps every source value, applies operator-approved definitions and source authority, records how each disagreement was resolved and by whom, and traces every governing figure back to its sources. A warehouse answers “what did each system say?” The operating record answers “which number counts, and on whose authority?”
Designed, not yet implemented in any community. The comparison describes two categories of infrastructure; it is not a production result.
Side by side
Different jobs, often confused for one.
Primary job
Warehouse: Consolidate data from many systems for reporting, analysis and history.
Operating record: Establish which value governs each decision, and keep the evidence for that determination.
When two systems disagree
Warehouse: Stores both values, or applies a transformation rule written by the data team. The disagreement is often invisible downstream.
Operating record: Keeps both values, surfaces the disagreement, and records the operator’s determination of which one governs, for which purpose.
Who owns the definitions
Warehouse: Usually encoded in SQL or a semantic layer maintained by data engineering.
Operating record: Written, versioned and approved by the operator’s accountable roles — finance, operations, clinical leadership.
Source authority
Warehouse: Implicit in which table or join a query uses.
Operating record: Explicit per field and per purpose: which system has the final word for staffing census, billing census or investor occupancy.
Lineage
Warehouse: Technical lineage — which pipeline moved which column.
Operating record: Decision lineage — the source values, the rule applied, the person who approved the governing value, when and why.
Typical owner
Warehouse: IT or data engineering.
Operating record: The operator’s executive roles, with IT as steward of the infrastructure.
What it feeds
Warehouse: Dashboards, ad-hoc analysis, data science.
Operating record: Reports, owner packages, recommendations and AI that need to read from a number with known authority.
One number, two treatments
Month-end occupancy at one community.
Hypothetical example with synthetic figures.
The clinical system shows 92 residents. Accounting bills 90. The CRM shows 91 of 100 units occupied. None is wrong: each counts a different event on a different clock.
In the warehouse
All three values land in tables. The owner report’s query joins to accounting, so it shows 90. The staffing dashboard joins to clinical, so it shows 92. Nobody sees that the two reports disagree until someone asks, and the reason lives in the query, not in a record anyone approved.
In the operating record
All three values are kept and the disagreement is flagged. Under operator-approved rules, billing census governs the owner report and clinical census governs staffing. The record shows each value, the rule applied, who approved it and when — so either report can be traced back on request.
How they work together
Not a replacement. A missing step.
Systems of record produce the data. A warehouse or lake can store and organize it. The operating record applies definitions, source authority and reconciliation, and records the determination. Reports, owner packages and AI then read from governed values instead of from whichever table a query happened to use. Coherence before Intelligence: settle which number counts before anything is built on top of it.
When a warehouse is enough
For some operators, it is the right answer on its own.
- Your systems rarely disagree on the figures leadership acts on, or one system already governs each of them.
- Your main need is historical analysis and trend reporting, not settling which number counts this month.
- Definitions are stable, written and owned, and your data team can show who approved each one.
- No owner, lender or regulator is asking you to show how a reported figure was determined.
Questions to ask your team
Six questions that show which one you need.
- When our clinical system and accounting report different census, where is the disagreement visible, and to whom?
- Who approved the definition of occupancy used in last quarter’s owner report, and where is that approval recorded?
- Can we trace a figure from two years ago to its source values and the rule that produced it, without a data-team request?
- If a definition changes, can we restate prior periods and show both versions side by side?
- Which decisions read from the warehouse directly, and which read from figures someone adjusted afterwards in a spreadsheet?
- Before AI explains a variance, does it know which of the conflicting values has authority?
Common questions
Do we have to replace our data warehouse?
No. A warehouse is a sound place to store and analyze history, and it can be one of the inputs an operating record reads from. The operating record adds the governance step the warehouse does not perform: deciding, with evidence, which value governs.
Isn’t this the same as master data management or a semantic layer?
They overlap. Master data management settles the identity of things — which resident, which community. A semantic layer standardizes calculations. An operating record also records source authority per purpose and the human determination when legitimate sources disagree, with decision lineage behind each governing value.
Can a warehouse be configured to do this?
Partly. Teams can add approval tables, versioned definitions and audit columns. The practical question is who maintains that apparatus as definitions, systems and portfolios change, and whether the business owners — not only the data team — control it. The page on designing your own governance architecture covers that choice.
Is SeniorCRE’s operating record available today?
The Operator-Controlled Operating Record is designed and not yet implemented in any community. Current capability states and production-entry gates are published at the Trust Center.
Where SeniorCRE fits
SeniorCRE is the operator-controlled operating infrastructure for senior housing & care. The design intent is that the operator keeps every authority decision while the record preserves the evidence. Deployment is the operator’s choice: keep the systems you run today and govern what they produce, or deploy SeniorCRE itself as the system of record for defined functions.
No operator-production deployment, live named connector, measured portfolio rollout or standard go-live duration is claimed. Replacement of a fixed number of systems within hours is not an available offer or a measured result. Medication administration/eMAR and third-party write-back are excluded from starting scope. Synthetic examples, architecture documents and controlled tests are not operator acceptance.
Bring one contested number to a working session.
We review how one of your figures is defined, sourced and approved today — not a product demonstration.