Research · Working Paper

Cash Custody Network Analytics

Supervisory-grade tail-risk analytics for delegated banknote custody networks.

Central banks delegate custody. Accountability stays central. The analytics must close that gap.

The gap

Delegated custody needs explicit tail governance.

The page is not a product claim. It is a public research surface for how SAA frames supervisory analytics when cash custody is delegated but accountability remains central.

01

Buffer ≠ constraint.

Inventory adequacy is set as a buffer over expected demand – not as an explicit limit on the probability and severity of service failure.

02

Coupled shocks.

Physical events raise cash demand and degrade delivery capacity at the same time, for the same reason. Treating them as separable understates required reserves.

03

Delegated execution, central accountability.

Predictive automation in reallocation widens the control gap unless it operates under fail-closed supervisory discipline.

Architecture

Four layers, one supervisory control surface.

L1 Network topologyNode concentration, effective network size, jump-to-outage exposure.
L2 Adequacy as tail constraintService failure bounded by probability and severity, not by buffer convention.
L3 Coupled stressDemand and capacity shocked jointly; marginals held fixed.
L4 Fail-closed supervisionNo output without controls cleared; aligned with prevailing model-risk and risk-data expectations.

Illustrative numbers

Synthetic parameters show the direction of the problem.

The numbers are deliberately bounded: they demonstrate mechanics, not operator performance.

10.21 → 9.50%

Stylised 12-node network: effective network size falls when delegated topology concentrates failure exposure.

× 2.6

Reserve independence assumption understates expected unserved demand when demand shock and delivery degradation are coupled.

Example

One event, two shocks.

A hurricane raises demand for physical cash and stops the trucks that would deliver it. Treating those as separate shocks understates failure severity in the synthetic example.

Twelve vaults. Effectively 10.2. Coupled stress returns the boundary, not a number the evidence cannot support.

Boundary. All results are illustrative, generated from stated synthetic parameters. No operator data are used. Operator-dependent quantities are identified and withheld, not estimated.

Founder proof

Written from inside cash operations, not from a whiteboard.

Written by the executive who ran a national cash and vault operation – 30+ centres, $50M+ daily – and kept it uninterrupted through the 2022 invasion of Ukraine.

Oleksii Slieptsov, CFA · Founder, SAA

Sealed unmarked note bundle
Cash custody fixture · public surface, no operator data

Working paper

From Buffer Heuristics to Governed Tail Risk

A supervisory analytics framework for delegated cash custody networks: topology, adequacy as a tail constraint, coupled stress and fail-closed supervision.

Research · SSRN Open working paper index Author page used until direct paper URL is published

Supervisory briefing

Bring one delegated custody question.

SAA can walk through the synthetic model, the withheld-operator-data boundary and the supervisory evidence structure before any real operator data is discussed.

Request a supervisory briefing →
Verify this page: Data room · Methodology
Scroll to Top