Capacity Is Not a Count: Staffed Public-Service Availability Under Changing Demand

A room is not yet a service

A shelter reports 300 spaces. A hospital reports 40 beds. A transit agency reports eight vehicles. A public assistance center reports ten service positions. Each number may be accurate and still overstate what the community can use.

The shelter may have staff for 220 spaces, accessible sanitation for fewer, and water through only the next shift. Beds may lack required personnel, equipment, or ancillary support. Vehicles may lack qualified drivers, fuel, accessible configuration, or an open route. Service positions may depend on trained staff, interpreters, identity systems, connectivity, or an accepting back office.

Capacity is therefore a conclusion about a service in time, not a count of physical objects. It changes when demand moves between windows, queues accumulate, a skill becomes unavailable, a resupply slips, or two sources disagree about status.

This paper separates two claims. Public guidance establishes that preparedness and response depend on current resource status, defined capabilities, qualified personnel, throughput, supporting services, accessibility, and responsible authority. Grid enters only as a proposition: represent selected aggregate relationships in an executable model, preserve uncertainty, and test the model in a bounded exercise. The cited agencies do not describe, endorse, or validate Grid.

What official guidance establishes

FEMA's NIMS Guideline for Resource Management Preparedness treats an inventory as more than a tally. A resource record includes status—availability and the time needed to prepare for deployment—and can bind to a resource typing definition or qualified position. FEMA's current National Resource Hub similarly covers personnel, equipment, teams, supplies, and facilities; its typing and qualification material establishes minimum criteria and a shared capability language. The Hub also states that it is not a deployment system and does not change mutual-aid authority. Preparedness information and authority to act are different things.

Health-care preparedness makes the capacity/capability distinction explicit. The current ASPR Hospital Preparedness Program guidance page describes coalitions coordinating across a defined area. ASPR's FAQ for its pre-decisional National Health Care Preparedness and Response Capabilities update identifies workforce, resources, patient movement, operational continuity, specialty care, and community integration. It calls the draft capabilities a reference, not federal requirements or cooperative-agreement guidance.

The 2024 NDMS Federal Coordinating Center Guide is even more concrete. It defines available beds as committed beds backed by supporting space, equipment, medical materiel, ancillary and support services, and staff. It calls for current bed reports and separately defines throughput as the estimated maximum number that can be received, off-loaded, staged, assessed, transported, and admitted in 24 hours. Transport resources, reception staffing, processing time, operating windows, reset time, personnel sustainability, and route or airfield constraints can limit throughput regardless of bed count.

For covered state and local emergency programs, accessibility cannot be treated as an optional modifier to an otherwise complete count. The Department of Justice's current emergency-planning guidance addresses accessible transportation, physical routes and service areas, shelter staffing, power and medication support, service animals, and effective communication. The page explains that state and local emergency-management programs must follow the ADA, while also noting that its guidance is informal and does not add duties beyond governing law. The responsible legal and accessibility authorities must assess the actual program.

FEMA's Community Lifelines connect government services, shelter, health, communications, and transportation. These sources support a planning premise: service counts require people, capabilities, dependencies, movement, and time. They do not prescribe Grid.

Four capacity layers

This four-layer frame is an evaluation device, not FEMA, ASPR, NDMS, or DOJ terminology. Evidence status—accepted, stale, conflicting, or unresolved—can apply at any layer.

• Nominal capacity is the published design, licensed, registered, or physical count: rooms, beds, vehicles, lines, desks, or daily slots.
• Staffed capacity is the portion supported by the required people and shifts, including qualifications, supervision, relief, and sustainable duty periods.
• Capable capacity is the staffed portion that also has the required configuration, equipment, supplies, support services, accessibility, connectivity, and governed service definition.
• Time-qualified capacity is capable capacity that is available for a named interval after preparation, travel, opening, turnover, reset, maintenance, resupply, and valid-through limits are applied.

These layers expose different constraints by setting. A shelter's sleeping area may exceed what sanitation, feeding, accessible routes, communications, volunteers, power, and water support. Physical beds may exceed staffed beds in the required service category. A fleet may exceed vehicles with drivers, fuel, maintenance status, configuration, route access, and cycle time. A benefits center, permit office, hotline, inspection team, or public counter may lack a trained role, language-access service, records system, security control, or accepting back office.

The distinction also prevents unlike units from being added. Ten general service positions are not ten specialist positions. A vehicle seat is not an accessible transport place. A cot is not a supported shelter service. A staffed hour in one period cannot cover demand that expires in another.

Demand arrives in windows and leaves queues

A planning model needs an aggregate demand profile, not only a daily total. For each window it can represent new demand, carried queue, service rate, reset time, supported queue, and an external deadline. Values may be exercise fixtures, observations, or estimates from another qualified process; Grid does not establish their predictive accuracy.

The core comparison is bounded: can the represented time-qualified service rate address the declared aggregate demand and carried queue inside the stated window? If not, the model can show a shortfall, an unresolved dependency, or the effect of a predefined contingency. It should not silently borrow tomorrow's capacity, assume that every nominal unit is interchangeable, or treat an average rate as valid during a shift gap.

This remains aggregate planning. The model does not decide who enters a shelter, who receives a bed, which rider takes a vehicle, whose case is heard first, or whether a person is eligible for a public program. Triage, diagnosis, clinical direction, individual priority, eligibility, allocation, legal interpretation, and protected-class determinations remain outside the proposition.

The Grid proposition

Grid proposes connecting the permitted aggregate planning facts as versioned dependencies rather than rebuilding them in separate rosters, capacity boards, transport trackers, supply sheets, and briefings.

A configured source can publish a named capacity, staff window, service category, queue, route state, support dependency, or resupply time. Each material value retains its source, definition, revision, observation time, valid-through time, and supplied uncertainty. Authored rules derive staffed, capable, and time-qualified states. When an accepted fact changes, affected conclusions recalculate; unrelated conclusions should remain stable.

The model can classify a predefined aggregate alternative as represented feasible, infeasible, conditional, or unresolved and explain the binding facts. A planner might see a shelter-service window, an operations leader a cross-site queue, and a decision owner the same revision's exceptions and required authority. Those are coordinated views of one modeled state, not proof that source systems integrate automatically. Every source binding, exchange, identity rule, timing rule, and failure behavior still requires configuration and testing.

The model then stops. It does not open a facility, summon personnel, move a vehicle, issue a benefit, deliver care, make a clinical decision, or exercise incident command. An external authority may accept, reject, revise, or defer a plan. A configured receipt can record that a named source update, authorization reference, or downstream message was acknowledged. That receipt proves only the recorded event—not understanding, execution, service delivery, compliance, or outcome.

Unknown is an evidence status

Capacity inputs often conflict without either source being obviously false. One roster may reflect the start of a shift while another reflects a call-out. A facility system may show open space after a logistics report's fuel window expired. A transport status may be current but use a different definition of accessible capacity. A resupply may be dispatched without a confirmed arrival.

The model should preserve those differences. Material inputs need, where applicable, a governed definition, source owner, revision, observed time, effective interval, valid-through rule, and quality or uncertainty state. A missing skill record, expired support fact, or unresolved definition conflict should make the affected capacity unresolved or unavailable according to an authored rule. Favorable imputation should never be the price of keeping a status green.

Sensitivity analysis can identify which assumption consumes the remaining margin. It cannot turn a stale fact into a current one. Sometimes the most responsible result is a refusal to conclude until a named owner refreshes or reconciles the evidence.

Fictional Exercise CIVIC-27

Consider a fictional aggregate exercise. Three shelter facilities have 620 nominal spaces: North 280, South 240, and East 100. At 16:00, staffing, sanitation, accessible-service support, power, water, and opening status leave North with 220 represented time-qualified units through midnight. South has 180 units through 22:00, for a combined North–South total of 400 during that interval; East's 100 remain conditional. The exercise demand profile is 360 abstract shelter-service units expected to remain through midnight. These units are planning counters, not people, households, clinical categories, or placement decisions.

At 17:20, an exercise inject adds 80 aggregate units arriving in the 20:00–22:00 window and remaining through midnight, bringing declared demand to 440. A support report confirms that South's generator fuel is valid only through 22:00. One 80-unit accessible shuttle block becomes unavailable. East's 100 units depend on an external activation, a staffed accessible-service plan, and a water delivery receipt.

Three predefined aggregate contingencies are evaluated. Keeping the original North–South plan has a 40-unit shortfall from 20:00 to 22:00 and a 220-unit shortfall afterward, when South's support expires. Counting East as immediately available is unsupported because East's required evidence and activation authority are absent, and it still leaves only 320 valid units after 22:00. A third contingency puts East behind three hard gates—activation authority, staffed accessible-service confirmation, and water receipt—adds a fourth gate for a South fuel receipt valid through midnight, and assigns a replacement 80-unit shuttle cycle by 21:30. Before the exercise injects satisfy those gates, the result is unresolved. After all four required gate records and the replacement transport window arrive, represented capacity is 500 against demand of 440 through midnight, leaving 60 units of margin. The fixture then classifies that contingency as feasible within its declared aggregate constraints.

The same exercise gives a next-day assistance center eight nominal service positions. Six are staffed, but a records-system outage and unavailable language-access support leave four capable positions for the four-hour morning window. At the fictional rate of 10 aggregate cases per position per hour, the center can process 160 cases by noon (4 × 10 × 4). An opening queue of 180 therefore leaves 20. The model exposes the represented throughput and queue; it does not decide eligibility, priority, case disposition, or service for any individual.

Nothing in CIVIC-27 is an operational result. The model has not created capacity, opened East, delivered water, restored a system, transported anyone, or served a case. “Feasible” means only that one fictional alternative passes the represented rules after its exercise evidence arrives. The Medical Plan Must Move shows the related capacity-and-time pattern in a medical-planning context, while Every System Works. The System Does Not. shows why separate current-looking views can still disagree.

Authority and assurance remain outside the count

Responsible organizations must define service categories, staffing and qualification rules, accessibility obligations, queue policies, support dependencies, and evidence freshness. The designated incident, facility, program, or jurisdictional authority retains activation and priority decisions under the applicable plan; Grid neither identifies nor replaces that authority. Clinical authorities retain diagnosis, triage, care, transfer, and patient-level decisions. Program and legal officials retain eligibility, benefits, due process, records, privacy, civil-rights, fiscal, contracting, and statutory interpretations. Resource owners and mutual-aid authorities retain release and deployment control.

Security, privacy, accreditation, and compliance also require evaluation of the complete operating environment. A model demonstration does not authorize a system, accredit a facility, validate a clinical process, establish ADA compliance, approve a records practice, or confer command or legal authority. The proposition makes no claim of a deployed public-service capacity product or measured service effect.

A bounded evaluation contract

Begin with one recurring aggregate capacity decision and its manual baseline. Name the source owners, capacity definitions, demand windows, staff and skill gates, equipment and support dependencies, transport and resupply rules, accessibility inputs, queue rule, external authority, and excluded individual decisions. Preserve the fixtures and calculate expected results independently.

Useful bounded measures include:

• accepted inputs with source, revision, definition, and valid time;
• stale, missing, duplicate, and conflicting facts detected under the fixture;
• time from an accepted changed fact to a stable revised aggregate status;
• declared demand windows with explicit feasible, infeasible, conditional, or unresolved results;
• throughput and queue calculations matching the independently calculated cases;
• capacity reductions with an inspectable staffing, skill, equipment, support, transport, resupply, accessibility, or timing reason;
• alternatives blocked when required evidence or external authority is absent; and
• receipts bound to the correct model revision, with delayed, duplicate, and replayed messages handled as declared.

Keep the evidence claim-specific. A conceptual argument describes the architecture. An illustrative case makes it concrete. Research-grounded evidence, as here, connects the problem to named public sources without validating Grid. A product demonstration requires inspectable behavior in declared fixtures. Customer-controlled evidence requires the responsible organization to control material cases, sources, environment, or operators. Independently validated evidence requires an appropriate independent method and stated boundary.

None of these metrics establishes prediction accuracy, appropriate service for an individual, legal compliance, clinical safety, deployment readiness, cost savings, or public outcome. Those claims need different authorities and evidence.

The practical next step is to map one capacity change through From a Changed Fact to Coordinated Action, define the evidence boundary with the decision evaluation worksheet, and compare the proposition with Grid Developers' explain-and-validate guidance.

A count says what exists on a list. Public-service availability says what qualified people and supported resources can deliver, accessibly, at a defined rate, in the window that demand requires—and which responsible authority still must decide.