Inventory Is Not Availability

The item on hand may still be unavailable

An inventory record answers a narrow question: does some system say that an item exists? An operational decision asks a harder one: can an eligible item reach a valid demand in usable condition, through an authorized path, before the need expires?

That difference is the space between inventory and availability.

A part can be on a ledger yet unavailable because its location is stale. It can be physically nearby yet unusable because condition or configuration is unknown. It can be serviceable yet committed to a higher-priority demand. It can be released yet unable to cross a closed corridor within the required window. In each case, the count is real and the mission value is not.

Contested logistics makes this relationship more volatile. Routes close, lift capacity changes, consumption departs from forecast, partner capacity moves, and communications become intermittent. A static report can therefore be accurate at its recorded time while the decision built from it is already wrong.

The institutional record supports the problem. Grid enters this paper only as a proposition: represent the facts, relationships, constraints, alternatives, evidence, and human authority in one executable model, then test whether that model improves a bounded logistics decision.

What the public record establishes

The Defense Department's logistics challenges are not reducible to one technology or one campaign. The public record shows several distinct failure modes: material that did not exist in sufficient quantity, material that existed but could not be located, cargo that could not be prioritized against constrained transport, and information that did not move coherently across organizations.

During Operation Iraqi Freedom, the Government Accountability Office documented hundreds of backlogged pallets and containers, inadequate asset visibility, duplicate requisitions, vehicle cannibalization when parts were unavailable or could not be found, and a $1.2 billion difference between materiel reported shipped to Army activities and materiel acknowledged as received. The report does not say every shortage was an information problem. It shows why the distinction matters: some supply was truly insufficient; some could not become operationally available because the system could not locate, describe, move, or account for it.

GAO's subsequent examination of critical items found that timely availability could fail through several interacting causes. GAO-05-275 identified inaccurate requirements and forecasts, delayed funding and acquisition, and ineffective distribution. Distribution itself depended on responsibilities, packaging, personnel, transport, and information systems. Availability was already a multi-variable decision, not a synonym for stock level.

The record also includes meaningful progress. In response to a recommendation in GAO-12-138, Defense Logistics Agency asset-visibility data was integrated into U.S. Transportation Command's IGC environment. GAO's implementation record says a single portal and web services eventually gave 7,500 users near-real-time access to shipment and stock information. Shared visibility mattered. It reduced the need to search separate views for the basic state of supply and movement.

But visibility is not the end of the decision. A portal can show a shipment without determining whether its evidence is fresh enough, its item is eligible for a demand, a feasible route remains, or a release authority has acted. GAO made this measurement point directly in a later review: asset-visibility initiatives needed performance measures linked to outcomes, not only evidence that a capability had been implemented.

The visibility problem also reaches upstream into the industrial base. In 2025, GAO reported that Defense Department efforts to understand foreign dependency were uncoordinated and limited in scope, leaving little insight into much of a supplier network that DOD estimated at more than 200,000 suppliers. The report recommended integrating and sharing supply-chain information, assigning responsibility, and testing ways to obtain missing country-of-origin data. This is a related but different layer of the same pattern: a risk model cannot calculate from facts that no authority collects or is willing to provide.

Current official demand signals point beyond tracking alone. Defense leaders have called for an integrated, resilient sustainment ecosystem and actionable data in contested logistics, while the Army's PORTAL SBIR topic asks for brigade-and-below software that can track, predict, recommend courses of action, ingest existing logistics data, run on laptops, and operate with limited connectivity. These are government requirements and priorities. They are not evidence that Grid satisfies them.

From a common picture to an executable availability model

An availability decision needs more than a consolidated count. At minimum, a bounded model should make these relationships explicit:

• Identity: Is this the correct item, variant, lot, configuration, or substitute?
• Location and time: Where was it observed, when, and how quickly does that evidence become stale?
• Condition: Is serviceability supported, disputed, unknown, or expired?
• Demand: Which unit, production line, facility, or mission needs it, at what priority and by what time?
• Eligibility and policy: What substitution, allocation, safety, fiscal, or release rules apply?
• Movement: Which routes, modes, handoffs, capacities, and time windows remain feasible?
• Consequence: What readiness, production, care, or operational commitment depends on arrival?
• Authority and provenance: Who supplied each fact, which model revision evaluated it, and who can authorize the external action?

The resulting object is not merely an inventory database. It is a decision topology connecting evidence to a proposed source-and-route match.

This distinction also clarifies interoperability. Sharing a field named quantity does not ensure that two organizations mean available quantity, serviceable quantity, releasable quantity, or quantity observed within an accepted freshness window. Technical connectivity can move a value while semantic differences preserve the wrong decision. The model must make the governed meaning and revision of exchanged values explicit.

The Grid proposition

Grid proposes that a logistics team author this bounded decision as a reactive computational model rather than reimplement it separately in a dashboard, spreadsheet, planning script, and briefing.

Configured connectors bring named source revisions into the model. Authored fields retain identity, location, condition, time, confidence, and provenance. A dependency graph traces which readiness or delivery conclusions depend on a changed fact. Predicates express hard eligibility and release conditions. Allocation, flow, routing, and scheduling methods evaluate represented alternatives within declared constraints. Explanation shows why an item or route was accepted, rejected, or left unresolved. Different sheets, maps, boards, documents, and APIs can present the same modeled revision for different roles.

That is an architecture claim, not a logistics outcome claim. It does not establish that an organization's source data is correct, that every domain rule has been modeled, or that a returned plan is globally optimal. It does not mean every external system updates automatically. Each connector, constraint, exchange binding, and downstream effect must be configured and tested in its operating context.

Consider a fictional bounded case. A mobility element needs one critical part inside six hours. Three systems report no stock. A fourth records two candidates in theater. One candidate has a stale location; the other has current location evidence but no accepted serviceability record. The old workflow issues another requisition.

In the proposed model, neither candidate becomes available simply because its quantity is positive. The first remains unresolved on location freshness. The second fails the authored condition gate until an authorized source provides evidence. When that evidence arrives, only the affected dependency path recalculates. The model evaluates represented allocation and route options, exposes the binding window and capacity constraints, and presents a feasible candidate for a logistics authority to review. It does not release the part, contract transport, or dispatch a vehicle.

That scenario is illustrative. The items, sources, timings, constraints, and result are authored fixtures, not an operational deployment or performance benchmark. The Supplies Were Already There visualizes the evidence-to-route pattern. Contested Logistics shows a route disruption, while Every System Works. The System Does Not. shows why operations and logistics cannot reconcile only in the next briefing.

Authority, security, and accreditation remain external boundaries

An executable model should make authority visible; it must not pretend to create authority.

Source owners remain responsible for identity, condition, classification, and release evidence. Commanders, logisticians, contracting officers, clinicians, or other recognized officials retain the authority assigned by policy and law. An external integration must authenticate actors, enforce a release gate, invoke an effect, and return a durable receipt when the workflow requires those controls. A predicate that says AUTHORIZED is inspectable model state, not authentication by itself.

Selective exchange is likewise not a cross-domain solution. A configured model can publish named results instead of an entire local dataset, but deployment authorities must still determine classification, releasability, need to know, privacy, transfer controls, and approved interfaces. Package signing can establish origin and integrity; it does not confer accreditation or make the package appropriate for every environment.

Grid makes no claim here of an existing authorization to operate, classified-domain accreditation, or compliance for a particular program. The applicable organization must assess the complete deployed system under its authorization process. For DoD systems, DoD Instruction 8510.01 establishes the Risk Management Framework; a successful model demonstration does not replace that process.

Local execution also does not prove full operation through denied, disrupted, intermittent, and limited communications. A local model may continue within validated local bounds, and configured deliveries may recover from a bounded interruption. Grid does not supply the communications path, guarantee indefinite offline synchronization, or resolve every two-way conflict after reconnection.

A bounded evaluation that can produce evidence

A credible evaluation begins with one recurring decision, not an enterprise transformation. Choose one item class, a small number of authoritative sources, one demand-priority policy, a represented route network, and one accountable release decision. Record the manual baseline before building the model.

Then run the same controlled changes through both workflows:

1. A location observation becomes stale.
2. Condition evidence is missing, contradicted, and later restored.
3. A duplicate demand enters under a different identifier.
4. Transport capacity falls or a route closes.
5. A higher-priority demand competes for the same eligible stock.
6. The selected option requires an authority the model does not possess.

Preserve the input fixtures, model and package revisions, evaluated alternatives, solver status, explanation, human decision evidence, and any downstream acknowledgment. Include missing and conflicting evidence on purpose; a demonstration that works only when every source is clean does not represent the institutional problem.

Measure the result in five groups:

• Evidence quality: verified-location rate, verified-condition rate, source freshness, unresolved contradictions, and demands with complete provenance.
• Decision flow: time from changed fact to revised availability state, manual rekey steps, duplicate requisition rate, backlog age, and time to a feasible source-and-route match.
• Model behavior: percentage of represented demands with explicit feasible, infeasible, or unresolved status; rejected options with an inspectable reason; and sensitivity to missing or delayed evidence.
• Coordination: conflicting revisions across views, time to logistics and command acknowledgment, and downstream effects with durable receipts where integrations exist.
• Human authority: recommendations reviewed by the correct role, exceptions requiring escalation, and actions refused when required evidence or authority is absent.

Do not convert those measures into promised savings or mission effects before the evaluation produces them. Product-demonstration evidence requires an inspectable Grid model with declared fixtures and limits. Operational claims require customer-controlled acceptance criteria, integration testing, security authorization, and observation in the intended environment.

The practical next step is to map one changed logistics fact through the From a Changed Fact to Coordinated Action journey, use the decision evaluation worksheet to define the baseline and authority boundary, and inspect the Grid Developers guidance for explaining and validating a changing decision.

Inventory is a fact about recorded stock. Availability is a conclusion about evidence, policy, movement, time, and authority. In contested logistics, that conclusion must remain alive as each of those conditions changes.