The annual service line dispute is not an arithmetic problem. It is a participation problem, and it has a structural solution.
There is a meeting that happens once a year in most health systems. Finance presents service line performance, clinical directors dispute the allocation basis, and everybody leaves having agreed nothing except that the exercise was frustrating.
This paper argues that the dispute is rarely about arithmetic and almost always about a model the clinical side had no part in building. It sets out why no allocation method is neutral, why that makes agreement rather than accuracy the objective, and what a data platform has to provide for the agreement to hold.
If you read nothing else, read these. The analysis that follows sets out the evidence for each.
Allocate theatre time by minutes and complex cases look expensive; allocate by case and short procedures look inefficient. Every method disadvantages somebody, which is why the model must be agreed rather than imposed.
The technically better allocation that clinical directors did not help design will be re-litigated every year, and the exercise produces frustration rather than decisions.
Drill-through from service line margin to the underlying transaction is what stops a challenge becoming a stalemate.
Retrofitting row-level security onto a published semantic layer is disruptive and it always happens at the worst possible moment.
Shared costs — theatres, imaging, pathology, overhead — have to be attributed somehow. Each method is defensible and each produces winners and losers among the clinical directors in the room.
When the model is produced by finance and presented to clinical leadership, the presentation is the first time the affected parties encounter the assumptions. The predictable response is to challenge the assumptions rather than to act on the result, and the challenge is legitimate because they were not consulted.
The organizations that have broken this cycle did one thing differently: they ran the allocation design as a joint workshop before any engineering started, and they wrote down the reasoning for each driver rather than only the driver itself.
Clinical volume sits in the electronic health record. Cost sits in finance. Payer performance sits in the revenue cycle system. Producing a joined view has historically meant a monthly extract exercise performed by one analyst, which is slow and, more importantly, unauditable.
Unifying these on a governed foundation changes both. The figure arrives continuously rather than monthly, and any published number can be traced back to source — which is what turns an internal report into something a board committee will act on.
The specific capabilities that matter are less exciting than the governance around them: an allocation reconciliation that proves total cost in equals total allocated out, a service line dimension built once rather than derived in report logic, and row-level security designed before publication.
Health system data is seasonal and case-mix sensitive, which makes short-period conclusions unreliable in a particular way: they are usually directionally plausible and quantitatively wrong.
We baseline against a period the organization has already closed and signed off, and we compare across a full year before drawing conclusions about a service line's trajectory. This is slower and it is the only version that survives a second board committee.
Every paper in this series ends with a framework you can run internally. We would rather you used it and reached your own conclusion than took ours on trust.
Five stages. The first is the critical path, not the preliminary.
Clinical and finance leadership together, before any engineering. Agree drivers per cost pool.
Write down the reasoning for each driver, not just the driver. It is what settles the challenge later.
Show the model against a period both sides already understand and have signed off.
Provide drill-through to the transaction, or the number will be disputed rather than used.
A named owner for the definitions and an annual review with a change process.
The same argument lands differently across an executive team. These are the three versions worth separating.
Microsoft's own documentation for the product behaviour described above. We would rather you verified the basis than accepted our summary of it.
On these references: each entry names a Microsoft Learn article or documentation area by title, because deep links change while titles are stable. Searching the title on learn.microsoft.com will reach the current version. Where we have cited a figure or a product behaviour, it is Microsoft's statement rather than ours; where we have given a number of our own it is labelled as such in the text.
We will build the cost model for one service line against your own data and walk your clinical and finance leads through it together, before anything wider is committed.
An examination of why AI deployments in provider organizations pause, and what the sequencing should be instead.
An estate designed for hosting usually needs rework before it can support AI workloads safely. This paper sets out what changes and why designing for it now is cheaper.
The question is not how long until systems are restored. It is how long the organization can deliver safe care without them.
Describe the situation in your own words.
