The Capital Slate — Hospital Network Planning

Where does the next unit
of capacity return most?

A ranked slate of candidate capacity additions, each one decomposed — beside a map that answers back. And the column nobody else in this market can compute: how much of that return sits in capacity a statute already removes on a published date.

California reference implementation · 458 licensed acute hospitals · 430 assertions green On-prem · no keyed provider · the runtime never calls a source host

© 2026 Tabaqat · Built on Strata — sovereign geospatial applications. The estate and the candidate slate are GENERATED (seed 20260826) and illustrative — not any operator's. A geographic screen, not a certification.

The decision this drives

Which capacity addition the board funds next.

Read at a capital committee, by a hospital-group strategy lead or CFO. The output is a ranked investment case with the internal trade-off named — not a facility dossier, and not a population count against a threshold. The rank order is the artefact the meeting argues over.

Which is why the rank order is the app: changing the sort renumbers the map, not just the list, because the rank lives on the record. A slate that reorders under a different reading of "return" is the point, not a glitch.

Buyer: hospital-group strategy lead / CFO Private operator — not the public "who is underserved" reader

17candidate additions on the shipped slate — all 17 on screen at first paint, no cap, no top-N

Ranked on demand won from competitors, rank 1 is a 48-bed unit at a site the group already runs — 21,744 of modelled annual demand, none of it taken from itself.

Re-sorted on the 2030-held share, a different candidate leads at 100 % — and it wins nothing at all. Same slate, same day.

The finding that shaped the app

In California, no rule gates adding capacity.
A hard rule gates the capacity that already exists.

Certificate of Need was repealed in 1987

There is no need-determination hearing to win in this market and no published bed-need formula an app could render as its governing rule. Roughly 35 states and DC still operate some form of CON; California is not one. A CON-shaped device here would model a rule that does not exist.

The rule that bites is a deadline

Health & Safety Code §130060: a building used for general acute care inpatient services after 1 January 2030 must meet the structural and nonstructural deadline requirements — or be retrofitted, replaced or converted to non-acute use.

So "the next unit" is often not growth

The board is not choosing between expanding and standing still. It is choosing between expanding and being smaller in 2030 than it is now. RAND costed the statewide mandate at $34–143 bn, averaging over $92 m per building, with 34 % of hospitals already in financial distress.

The brief called this a siting problem. It is a replacement problem wearing a siting costume — and that is what put a statutory column on every card.

The market, page-verified

Two industries hold one decision. They do not share a map.

Product	What it can say	What it cannot say
Sg2 / Vizient	The most authoritative demand curve in the market, by service line, out ten years	A forecast and an advisory relationship — not a ranked slate of places
Esri Business Analyst	Real site selection — drive-time trade areas, Huff allocation, Measure Cannibalization	That tool is geometric overlap only — no demand is allocated. And no regulatory state
Caliper Maptitude for Health Care	5,428 hospitals with type, network and beds, plus HRSA service areas	A desktop mapping product with data attached — no ranking, no decomposition
Placer.ai · Trella · Definitive	Visits observed, referrals from real claims, the most complete provider directory	Footfall, claims and firmographics — no bed, no licence, no walls
HCAI's own seismic dashboard	The authoritative picture of which buildings are rated what	The state of an operator's walls — never the consequence for its market
The Capital Slate	Modelled demand, the internal transfer, and the published statutory fate of the buildings behind it — the join nobody else holds both sides of	Nothing about the 2030 term outside California — that rule has no analogue elsewhere

Search for a product binding seismic compliance to capacity planning and you get construction cost consultants, structural engineers and real-estate advisors — and no health-market analytics vendor at all. The market analysts model demand and never look at the walls; the construction advisors cost the walls and never look at the demand.

How a candidate's return is decomposed

Three terms — and only one of them is not modelled at all.

Won from competitors

Modelled demand currently allocated to another operator's site. This is growth. Every benchmark in the category does some version of it.

Transferred from our own sites

Modelled demand currently allocated to a site this group already runs. An internal reshuffle wearing a growth costume — and honestly, not unique: Huff allocation has decomposed capture this way for years.

Held in 2030-rated capacity

NEW — and the least modelled term in the table. Demand sitting in a building whose published SPC rating is not 2030-compliant. The rating, the date and the services are published facts about buildings, not inferences about patients.

The correction the build forced: it does not split three ways.

Displacement is a property of the incumbent building; ownership is a property of the incumbent operator. So every displaced unit is also either won or transferred — drawn as three segments it double-counts. It ships as two segments (won + transferred = captured, asserted exact to 1e-6) plus a cross-cutting hatch band for the 2030 share, on its own rail.

mostly wonthis candidate is growth

mostly transferredan internal reshuffle — the number a committee argues over

mostly 2030-heldnot an addition at all — a replacement the statute already scheduled

The application, first paint

Everything needed to defend the order is on screen.

The Capital Slate at first paint: the ranked rail on the left carrying candidate #1 with its decomposition bar, the California map on the right with numbered candidate dots, the legend counting each class, the KPI block at the foot of the rail and the notice bar beneath.

Shipped build, 2026-08-26 — a browser screenshot taken by the automated driver, not a mock-up. Dark is the working console; light is the committee mode and what the export prints as.

The rule chip, in the header. The allocation rule, the demand lens, the travel mode and the origin policy — printed above every number they produced.
A persistent notice bar. The generated estate, the SPC vintage, the case-mix coverage, the screening-only scope — never the status line, which is transient.
The rail, ranked and renumbered. Each card carries won, transferred, the 2030-held share, the captured total, and the buildings behind that figure, ordered by contribution.
The legend filters, and counts. Click hides a class, shift-click isolates, Esc clears — features gone, not faded. Rows keep n of N.
The unmet chip. 2.8 % of modelled demand fits no site at this rule. Reported, never redistributed — force-placing overflow erases geography from the answer.

The number that is new

The 2030-held share, per candidate.

What fraction of a candidate's captured demand sits in capacity that state law removes on a published date — with the buildings behind it named, ordered by demand actually donated.

candidate New address — Menifee city, Hemet city ▨ GENERATED
won from competitors 0 · transferred from our own sites 11,380
= captured demand, per year 11,380
held in 2030-rated capacity 100.0 %
behind it: Perm-ID-10541 · HEMET GLOBAL MEDICAL CENTER · 1 SPC-1 building
rated 2021-12-29 · SPC-1 subset only · NPC not published to this app

It turns a market question into a capital question with a deadline on it. A candidate whose demand is mostly won is growth. Mostly transferred, and it is an internal reshuffle. Mostly 2030-held — and it is not an addition at all.

And the app says who decides: HCAI. It renders a published rating and a statutory date. It never asserts that a building will close.

The rail17 of 17 · sorted by 2030-held share

1New address — Menifee city100.0 % won 0 · transferred 11,380 · 25 proposed beds
1 SPC-1 building behind the whole figure

2HEMET VALLEY HEALTH CARE CENTER100.0 % won 0 · transferred 5,003 · 11 proposed beds

3SHRINERS HOSPITALS FOR CHILDREN99.9 % won 5,458 · transferred 0 — pure growth, and pure replacement

4New address — Murrieta city… … 13 more · the rail scrolls · nothing truncated

15.6 % of the estate's own beds sit in 2030-rated capacity · 327 of 2,093 ▨ GENERATED

Rank 3 is the case that makes the column worth drawing: maximum growth and maximum statutory exposure at once. No single ratio would have separated it from rank 1.

The sentence that governs every decision below it

The rank order is the app.

The same slate re-sorted by 2030-held share: rank 1 is now a new address at Menifee city with a full amber transferred bar, a red hatched 2030 band beneath it, and the building behind the figure named — Perm-ID-10541, Hemet Global Medical Center, one SPC-1 building.

One change of sort. A different candidate at rank 1, a different decomposition, and the numerals on the map renumber with it — because RANK lives on the record and the source is re-set on every change.

Four sorts, each a different reading of "return": won from competitors · total captured · 2030-held share · transferred from our own sites.

The demand lens — four published MS-DRG families, each paired with a published CDC PLACES measure. It re-weights, renumbers and recolours in one frame.
Scope — county and emergency level compose one WHERE, round-tripping as ?where=…: a slate can be sent as a link.
The pin tray — pin two candidates and the combined internal transfer recomputes. That is the number a committee argues over when two candidates overlap.

47 numbered behaviours, each mapped to the suite that exercises it

The signature interaction

Bidirectional, with a visible release.

Click a rank card and the map flies to that candidate, pulses it and opens its popup. Click a numbered dot and the rail scrolls that card into view and pulses it. Click either again and everything releases.

The outbound half — row to map — is what every viewer in this shape ships. The inbound half is the one they skip, and without it the map is a picture rather than a partner.

Arrow keys step the rank, adopting as they go, clamped at both ends.
Esc unwinds one thing at a time — drawer, then isolate, then hidden, then selection, then pins, then scope.
Every pan recomputes "In view", keeping its denominator.
Export is a one-page landscape slate sheet with light tokens forced, plus a CSV carrying the class words, the rule chip and every vintage.

Proof, not promises

Built, driven and measured — 2026-08-26.

430assertions green, 0 failed — 108 live · 235 offline · 87 in real headless Chrome

47behaviours specified, each mapped to the suite that exercises it — against a floor of 3

4.67worst informational contrast ratio, in both light and dark — the floor is 4.5

17of 17 candidates on screen at first paint — no cap, no top-N, the whole slate

The join is the product, so the join is asserted

The CCN bridge joins zero rows naively — one publisher, two spellings. It resolves 268 of 272 after zero-padding; the four residuals are carried as unresolved, never dropped. The SPC join resolves 40 of 43; the three that left the licence register keep their own class.

Both "keyless" default basemaps were serving a placeholder at HTTP 200

One returns the real map with "API KEY REQUIRED" composited over every tile; the other a fixed "418 · Access blocked" image. Both pass a URL check, a !key= check and a naturalWidth>1 check. A screenshot is what caught it.

Three defects had both non-visual suites green

The rank numerals — the app's whole signature — validated, added, threw nothing and drew nothing. Fixing that broke the entire layer silently: a data-driven icon-image fails in the tile worker with a clean console. Now rasterised to canvas icons, one symbol layer per class.

Reading the figures: the supply register, the seismic ratings, the tract geography, the prevalence estimates and the case mix are real, published and re-probed on 2026-08-26. The estate and the candidate slate are GENERATED (seed 20260826) under four conditions that all hold — and are labelled ▨ GENERATED on every card, in the legend, in the KPI block and in every export. No generated figure is ever summed into a measured one.

Stated as boundaries, not caveats

What this application will not do.

No patient-level data, under any framing. Not de-identified, not sampled, not for a demo. Transfer is a modelled allocation under a stated rule — never observed patient movement, and the app is not permitted to phrase it as one.
Not a compliance determination. It reports a published rating and a statutory date. Retrofits complete, plans get approved, delays get granted — HCAI decides all three.
Not a money model. No rent, no footfall, no payback, no capital cost. Those lanes render EMPTY with their citation, and no sort touches a money figure — including the three payment columns the case-mix file does publish.
Read-only, end to end. No write path, no editing, no case management — so it runs identically on Strata Serve and on ArcGIS.

Travel is a labelled straight line, car assumed — it under-states travel wherever a mountain, a bay or a river intervenes. The error applies uniformly, so the order is far more robust than any absolute. A deployment names a self-hosted router; a public one receiving your estate geography is a security finding.
Every total from the case mix is a floor. It covers 272 of 458 hospitals (59.4 %) and suppresses by row omission below 11 discharges. Coverage gaps read "no estimate", never 0.
The 2030 term is a floor, not a measure. The NPC half of the statutory test is not published to this app — the complete file returns 403 outside California. Every card says so.
The 2030 rule is Californian and has no analogue elsewhere. The frame ports — rank, decompose, name the internal trade-off. Any deployment outside California must name its own rule in that slot, or ship without the term.

Every one of these is written into the delivered application's own README. A tool that hides its edges costs you the project it cannot finish.

Point it at your estate

One facility register.
One candidate list.

Your real estate in place of the generated one, your candidate additions with their proposed bed counts, and the rule that gates capacity in your jurisdiction. Two weeks, and this slate ranks your capital plan instead of a demonstration.

Swappable by configuration, not by code: the supply register, the demand geography, the prevalence lens and the statutory column. The three-term decomposition is a typology the app renders — not a list of California numbers it ships.

tabaqat.net → Solutions → Health & Community info@tabaqat.net

© 2026 Tabaqat · Built on Strata. Reference implementation over California; estate and candidate slate GENERATED (seed 20260826). A geographic screen — not a certification, not a compliance determination, and not a business case of record.