Unit-Cost & Breach Computability

In July 2024, the Department of Defense announced a unit-cost breach[1] in the Sentinel program. The program's per-unit cost grew from an original baseline of $118 million to $214 million—an 81.3% increase.[2] This exceeded the 50% Nunn-McCurdy statutory threshold for a critical breach. Total program cost reached $140.9 billion.[3]

In February 2026, the Air Force restructured the program,[2] introducing 450 new modular silos and approximately 5,000 miles of fiber-optic infrastructure.[4] A prototype broke ground at Promontory, Utah. These changes reset baseline computations, but without UC-BCK, the same gaps persist.

10 U.S.C. § 4371[5] establishes two metrics for breach detection:

The FY 2006 NDAA[6] introduced dual-baseline tracking, requiring programs to report growth against both the current baseline and the original Milestone B baseline—closing the "rebaselining loophole" that previously let programs reset their breach counters:

Source: 10 U.S.C. § 4371;[5] DAU Statutory Program Breach Definitions.[7]

Upon a critical breach, 10 U.S.C. § 4376[8] requires the Secretary of Defense to certify five things within 60 days—or the program is terminated:

1. The program is essential to national security.
2. No less costly alternative exists to achieve the required capability.
3. New cost estimates are reasonable (as determined by the Director of CAPE).
4. The program has higher priority than programs whose funding would be reduced.
5. The management structure is adequate to manage costs.

For Sentinel, USD(A&S) certified the program on all five criteria in July 2024.[1] Milestone B approval was rescinded, requiring restructuring and a new milestone review.

Sentinel's breach was not "impossible to foresee"—it was impossible to compute reproducibly until seven months after cost overruns had metastasized. Three structural gaps explain why:

Gap MC1: Unit Definition Ambiguity

What is a "unit" in Sentinel? The CRS has documented recurring unit definition disputes[13] across historical programs:

• LCS: In May 2022, the Navy cancelled the ASW mission package, triggering a 37.3% PAUC increase over the original baseline[14][15]—not because remaining packages cost more, but because fixed costs were amortized over fewer units.
• BMDS: Continuous disputes over whether a "unit" is a single radar, a site, or a regional defense posture.[9]
• F-35: Recurring questions about unit scope across multiple annual assessments.[16]

Without a locked, machine-readable definition approved by the milestone authority, contractors and programs can redefine the unit retroactively, masking breaches or shifting them to different metrics.

Gap CS1: Baseline Comparability Masking

A 2005 GAO audit found that programmatic adjustments masked approximately 50% of Nunn-McCurdy breaches[17] between 2001 and 2003. GAO identified 25 programs with cost growth, but only 17 were formally reported as breaches.

• RAH-66 Comanche: GAO documented 62% PAUC growth over 27 months,[17] masked by excluding unit cost increases associated with a restructure and a reduction of 561 helicopters. The underlying cause was failure to mature composite materials and avionics software.

Without a frozen, auditable baseline-to-current crosswalk, breaches can be masked indefinitely.

Gap CS2: Computational Non-Reproducibility

Even when PAUC and APUC figures are published, no independent party can reproduce them from source data. The calculation depends on:

• Scope of PAUC numerator (the statute specifies RDT&E + Procurement + MILCON—but which cost elements map to which category?)
• Definition of "total quantity" (originally planned, currently planned, or realized?)
• Allocation of indirect costs to units
• Treatment of quantity adjustments mid-program
• Which baseline is the reference (original approved, last updated, or custom?)

Contractors submit cost data to DAVE,[10] but those data are optimized for narrative rather than reproducible computation. CADE holds more detailed FlexFile repositories,[12] but the mapping from FlexFile categories to Nunn-McCurdy unit cost is manual, undocumented, and varies by program.

Cost growth in complex acquisition follows a feedback loop. Cooper (1993) documented how rework cycles in development projects compound through iterative discovery and correction.[18] Ford and Sterman (2003) modeled how rework concealment in concurrent development creates cascading cost surprises.[19] Gansler (1999) termed this "funding migration," where rising O&S costs for aging systems consume modernization budgets, forcing quantity reductions on new programs.[20]

F-22 Raptor

The F-22 program initially planned 750 aircraft.[21] Successive budget cuts triggered quantity reductions:

Total RDT&E reached approximately $24.3 billion,[22] making per-aircraft development cost unsustainably high.

DDG-1000 Destroyer

The DDG-1000 was originally planned for 32 ships.[23] Cost overruns led to successive reductions:

GAO documented cost growth driving down quantity,[24] yet each reduction was justified as a "force structure trade" rather than programmatic failure.

JDAM: The Positive Case

The JDAM achieved a 25% unit cost reduction and procured more than double the originally planned quantity.[16] Unadjusted total cost growth substantially overstated the program's cost management problems, because quantity increases drove total spending up even as per-unit costs fell. Both directions matter—adjustments can hide bad news or obscure good news.

In January 2024, the Air Force notified Congress of the breach.[3] By July 2024, the Department formally announced the results: unit cost had grown 81%.[1]

Three Cost Baselines, Three Different Stories

The 81% figure refers to PAUC growth ($118M → $214M per unit) and is valid as a ratio. As of March 2026, no new cost estimate reflecting the February 2026 restructure has been published.[2]

Where Did the Cost Growth Come From?

The Department's root cause analysis determined that approximately 80% of cost growth originated in the Command & Launch segment[1]—ground infrastructure, real estate, and fiber-optic communications rather than the missile airframe.

However—and this is critical—no public segment-level dollar breakdowns exist. Only the aggregate 80% figure is acknowledged. This opacity is precisely the computational non-reproducibility that Gap CS2 describes.

The February 2026 Restructure

• Modular Silo Design: 450 new silos using modular construction.[4]
• Fiber-Optic Network: Approximately 5,000 miles of hardened fiber-optic cable.
• Prototype at Promontory, Utah: Full-scale prototype validating modular design.
• Extended Schedule: IOC targeted for the early 2030s.[26]

These changes reset the baseline, but without UC-BCK, they also reset the reproducibility clock.

How the Three Gaps Manifested

MC1 (Unit Definition): Sentinel was costed as a "missile system." Requirements crept in for integrated command, control, communications, and cyber-hardening. The program used baseline adjustments to avoid reporting breaches, until the 81% threshold could no longer be suppressed.

CS1 (Baseline Masking): Adjustments for "threat environment changes" and "STRATCOMM integration" were justified in narrative, but no formal crosswalk documented "original scope" vs. "new scope."

CS2 (Non-Reproducibility): The 81% figure appeared in press releases and SARs, but the underlying calculation was not published. Congress and GAO had to accept the DoD's announcement or fund an audit.

Two systems hold the raw data needed: DAVE[10] and CADE.[12]

DAVE: System of Record

DAVE is the DoD's authoritative system for acquisition program data,[10] API-oriented[11] but access-restricted. It presents aggregated views optimized for narrative rather than reproducible computation.

CADE: Evidence Library

CADE[12] houses FlexFile repositories in the CSDR format,[27] historical baselines, and cross-organizational access for audit.

The critical gap is not data availability—it is data integration and reproducibility. DAVE and CADE are disconnected. No party can independently reproduce a breach computation from first principles.

Legal Infrastructure Exists

• MIL-STD-881F[31] defines WBS standards for consistent functional decomposition.
• 10 U.S.C. § 3221[32] establishes CAPE's authority over cost estimation policy. § 3223[33] mandates CAPE's access to records and data including contractor data. § 3227[34] requires CAPE to develop collection methods for quality cost data.
• DFARS 234.71[27] establishes CSDR requirements for consistent reporting formats.

What is missing is the governance and architecture for making that data reproducible and contestable. UC-BCK bridges that gap.

UC-BCK is a governance and technical architecture closing the three gaps. It integrates existing systems (DAVE, CADE, FlexFile) with three new capabilities: a locked, machine-readable unit definition; a reproducible computation engine; and a multi-view evidence ledger.

Design Principles

11 Components

Multi-View Computation

If quantity had been reduced (as in F-22 or DDG-1000), these views would diverge—and that divergence is precisely the information Congress needs.

4-Phase Roadmap

Technical Feasibility

All components are TRL 9. DLA's ASSR system demonstrates the approach within DoD: 98% manual work reduction on the Pega platform, processing 250,000–300,000 transactions annually.[35] DAVE is already API-oriented.[11] DFARS 234.71 already requires audit trails.[27] Computation completes in under 1 second per program.

Risk Assessment

Organizational (HIGH): Program managers will resist transparency. Mitigation: pilot with Sentinel post-restructure, where the Air Force has committed to transparency.

Technical (LOW): All proven technology. Mitigation: iterative pilot approach.

Political (MEDIUM): Contractor lobbying against mandates. Mitigation: frame as transparency tool benefiting all parties.

Stakeholder Implications

Congress: Earlier breach detection. Independent verification. GAO: Automated replay; reduced audit cost.[36] Program Managers: Early warning; defensible reporting. Contractors: Clearer unit definitions; formal appeal mechanism. Researchers: Standardized, reproducible cost data for independent analysis.

Conclusion

UC-BCK closes three gaps that have persisted for four decades. It produces a cost accounting architecture that is reproducible (any party can verify any breach determination), auditable (every figure has a source and trail), contestable (disagreements routed through governance, not suppressed), and early-warning (continuous monitoring before breach thresholds). It requires legislative updates aligned with the FY 2006 NDAA framework[6] but no new authorities or systems. Achievable within 5 years.