Transaction-Level Risk Accounting in Banking: Payment Orders Explained

Home › Knowledge Base › Transaction-Level Risk Accounting in Banking
Risk Accounting Knowledge Base • Banking Application

Transaction-level Risk Accounting links banking activity, structural exposure and assessed controls to a traceable non-financial-risk measurement process. In the published method, daily product activity selects a value weighting, which combines with exposure factors and mitigation to produce Risk Units. Transaction records support attribution, but do not automatically imply a distinct RU for each payment.

A bank may process millions of payments, securities instructions, loans and customer transactions. These activities create operational demands long before any loss is recorded. A payment may be validated, screened, routed, reconciled and settled through several business components and systems; a failed step can produce a loss, a service disruption or a regulatory issue. Yet a monetary incident database records only events that have been detected, classified and sufficiently quantified.

A current-risk management question is therefore different from a loss-history question: what exposure is the bank carrying as its processes accept and handle new business, and what part of that exposure remains after assessed mitigation?

Risk Accounting addresses that question through a measurement chain linking business activity, structural exposure and control assessment. The chain is designed to support classification by product and, where the method and source records support it, analysis by business component, customer, location and other organisational dimensions. [1, pp. 136–141]

The expression transaction-level needs precision. The published worked example does not generate a unique, freestanding RU figure for each individual transfer. It starts with a daily-new-business measure by product, selects an approved Value-Based Weighting (VBW), then calculates and aggregates exposure by relevant risk families and business components. More granular attribution demands documented records and valid allocation rules; a simple pro rata division of product RUs by the number of payments would not by itself be a governed method result.

2. How the measurement chain works for banking transactions

The published banking framework distinguishes two kinds of inputs to inherent exposure:

Structural exposure. A product or activity is assigned an applicable Exposure Uncertainty Factor (EUF) for each principal risk family. For general processing, Hughes illustrates assignment using operational touchpoints; different families have different published criteria. EUF is a method-defined factor, not a probability of fraud, systems failure or loss. [1, pp. 138–139, 156–160]

Business scale. The institution records daily new business using the definition and time basis for its method version. The value selects a VBW from a discrete value table. A payment's monetary amount is not itself an RU, and business volume is not multiplied directly by EUF. [1, pp. 139–140, 161]

For principal risk family r:

Inherent RUr = EUFr × VBW

To assess mitigation, the published method maps relevant operating components to Enhanced Risk and Control Self-Assessments (E-RCSAs), using specified benchmarks, category weights and evidence of the operating state of controls. At the validated scope:

RMI (%) = 100 × ASAMSA

where ASA is Actual Score Aggregate and MSA is Maximum Score Aggregate, with a valid nonzero denominator. The residual exposure at that same scope is:

Residual RU = Inherent RU × 1 − RMI (%)100

Parent-level RMIs must be recomputed from the relevant ASA and MSA aggregates, not obtained by taking an unweighted average of departmental percentages. The computation must also respect approved product mappings, risk-family applicability, reporting dates and consistent aggregation rules. [1, pp. 140, 164–168]

Published Fictitious Case • Hughes (2023)

3. Worked example: Payment Orders in Hughes's published banking case

The following values are taken from Hughes's published Payment Orders example, whose inputs are explicitly fictitious and simplified. This case explains method mechanics; it is not a claim about the operations of a real bank. Importantly, it is not the same dataset as RASB's later Payment Orders controlled synthetic research demonstrator. [1, pp. 144, 148–149, 163, 166]

Step 1 — Identify the product and reporting basis

The published Payment Orders product is assigned US85 million daily new business. Under the published discrete Value Table, the applicable band is US64 million to under US$128 million, giving VBW = 27.0.

This calculation is based on the relevant daily product total, not on multiplying the number of payment orders by a guessed per-payment risk score.

Step 2 — Apply the published EUF assignments

The applicable principal families for Payment Orders are:

Principal risk family Published EUF Applicable VBW Inherent risk
Processing 5 27.0 135.0 RU
Selling 2 27.0 54.0 RU
Total for Payment Orders 7 — 189.0 RU

The book's other principal families are not assigned an EUF in this particular Payment Orders example. They should not automatically be interpreted as universally irrelevant to every payment product, institution or implementation. [1, pp. 148–149, 163]

Step 3 — Use the matching product-level mitigation aggregates

The published whole-product Payment Orders report gives:

  • Actual Score Aggregate (ASA): 7,854,030
  • Maximum Score Aggregate (MSA): 10,864,800

The product RMI is therefore:

RMI = 100 × 7,854,03010,864,800 = 72.2888% (approx. 72.3%)

The published summary rounds this to 72.3%. The score aggregates are published outputs of the worked example's component/assessment calculations, not an invitation to invent replacement control scores. [1, p. 166]

Step 4 — Determine residual exposure

Using the unrounded ratio:

RMI = 100 × 7,854,03010,864,800 = 72.2888% (approx. 72.3%)

Thus the summary reports approximately 189 inherent RUs, 72.3% mitigation and 52 residual RUs. This is a method-defined exposure result; it is not 52 failed payments, US$52 of expected loss, a fraud probability, or a regulatory capital amount.

Step 5 — Interpret a changed operating state

If payment activity grows, the valid daily-new-business measure and resulting VBW may change. If the quality of assessed controls changes, the applicable E-RCSA and RMI may change. These effects can be investigated separately instead of compressing the whole situation into one subjective risk category.

However, the discrete VBW schedule is not continuously proportional: an increase in activity that remains within the same approved band does not, on that input alone, change the VBW. Nor does merely asserting that a control failed establish a new score. A new residual RU result requires valid, governed inputs for the relevant reporting date and scope.

4. What makes transaction-level attribution auditable?

Transaction data can help provide the provenance of the business activity that feeds the measurement. That is useful even when the calculation runs by day and product rather than per instruction. A defensible implementation should retain or be able to reconstruct the following chain:

Data layer Examples of records What must be verifiable
Business activity Payment order ID, date, product, amount, value date, accounting and booking references Source completeness; inclusion and exclusion rules; duplicate handling
Classification Product catalogue, legal entity, originating business component, applicable principal risk families Effective-dated mappings and approvals
Daily new business Defined daily product total and supported currency normalisation Reconciliation to transaction/accounting systems and period basis
Structural exposure Approved EUF assignment, version and rationale Source criteria and change authority
Value weighting Value-band table, selected band and VBW Correct threshold and table version
Mitigation E-RCSA responses, benchmark evidence, categories, weights and effective dates Evidence, reviewer and status at the relevant date
Aggregation ASA, MSA, parent/child mapping and computed RMI Nonzero valid denominator and coherent reporting scope
Result Inherent and residual RU, calculation version, as-of date Reproducibility, change history and exception handling

A transaction or event ID alone does not make the computed risk position reliable. Completeness of the source feed, stable definitions, control assessment quality, aggregation logic and independent challenge are necessary. Likewise, posting a daily RU measure does not by itself mean that each payment carries a live risk price at the instant of processing.

5. The current RASB Payment Orders research: what it does and does not show

RASB's public Payment Orders Risk-Exposure Study (2025–2026) presents a controlled synthetic research demonstrator, identified as EXP-02 / Study-2. The public page describes 522 daily observations over 2025–2026 and eight quarterly assessments, with one governed primary synthetic measurement chain. [2]

The site offers three distinct ways of examining the same synthetic case:

  • A public Mini Lab for understanding the relationship between business activity, inherent risk, mitigation and residual risk through baseline views and hypothetical adjustments.
  • A public Management Dashboard for exploring monthly changes and questions relevant to operations, risk management, finance and executive leadership.
  • A restricted Model Lab, described as a fuller technical and academic instrument, where access requires registration or approval. Its existence should not be represented as unrestricted public access.

Recorded synthetic states and exploratory what-if adjustments are not the same evidence. The 522-day controlled timeline is valuable for testing calculation mechanics, reproducibility, sensitivity and longitudinal behaviour under its protocol. It does not independently establish that an RU predicts real bank fraud losses, offers a proven money-equivalent valuation, or outperforms existing methodologies in live institutional use. [2]

The 2023 book example (US$85 million, 189 inherent RUs and about 52 residual RUs) and the 2025–2026 synthetic research dataset are separate examples. Their inputs, versions and outcomes must not be merged or treated as observations from the same bank.

6. Conventional approaches, governance and regulatory context

Criterion Conventional operational-risk tools Risk Accounting application
Transaction records Processing, settlement, accounting or incident evidence Potential traceable source for daily activity and attribution
Loss database Past detected, reported and quantified incidents Complements current exposure measurement; can support later empirical calibration research
RCSA Assessment and prioritisation of risks and controls E-RCSA provides structured control-state inputs for RMI under approved benchmark rules
Risk scenarios / FAIR Hypothetical future losses and distributions in monetary terms Different output: current method-defined exposure in RUs
Capital formula Binding jurisdiction-specific prudential computation RU does not replace regulatory Pillar 1 operational-risk capital requirements

The Basel Committee's operational-risk management principles emphasize governance, assessment, monitoring, controls and internal reporting; BCBS 239 addresses robust risk data aggregation and reporting. These are relevant governance considerations, not legal endorsements of the RU method. Under the EU CRR3 framework applicable from January 2025, operational-risk own-funds requirements are calculated using the Business Indicator Component, not a residual RU ledger. [3][4][5]

Implementation questions for a bank include: Which activity is represented? At what time basis? Can daily business be reconciled to authoritative records? Are product and controls mappings effective-dated? How are control gaps, missing values and revised assessments treated? Who independently confirms the model version and the results? A calculation should not substitute a missing or invalid input with an arbitrary plausible value.

Operational value is a hypothesis that can be tested through controlled pilots and prospective observation. A stronger research programme would predefine measures of predictive utility, decision usefulness, generalisability, stability, comparison methods and out-of-sample evaluation. Mechanical reproduction is one essential condition, but it is not itself a complete empirical validation.

7. Frequently asked questions

Does Risk Accounting calculate a separate Risk Unit for every payment?

Not automatically. The published banking example uses product-level daily new business, approved exposure factors and control assessments. Per-transaction allocation would need an explicitly governed and validated method rather than simple division of a daily total.

Under the Risk Accounting method, an institution can calculate a current method-defined exposure from governed activity and mitigation inputs. This is distinct from establishing the probability or amount of a later financial loss.

Is the Payment Orders research based on actual commercial-bank transactions?

No. The current publicly described 2025–2026 research case is controlled and synthetic, with 522 daily observations and eight quarterly assessments. Its results demonstrate properties of that controlled case, not observed bank-loss predictive performance. [2]

Are Risk Units the same as operational-risk capital or expected financial loss?

No. Risk Units measure method-defined inherent and residual exposure. Monetary expected loss requires separate validated calibration; accounting recognition and prudential capital remain distinct questions under their applicable rules.

8. Continue exploring the method

Read the basic definition of the Risk Unit and the calculations for Exposure Uncertainty Factor, Value-Based Weighting, Enhanced RCSA and the Risk Mitigation Index. For different measurement objectives, see Risk Accounting vs ERM and FAIR.

Explore the actual Payment Orders research demonstrators and The Method. Researchers interested in independent testing or institutional pilots can use Collaborate & Contact.

Return to the Risk Accounting Knowledge Base for all nine topics.

Sources

  1. Hughes, Peter J. Risk Accounting: The Complete Guide to Quantifying and Accounting for Non-Financial Risks (Grosvenor House Publishing, 2023), especially pp. 136–141, 144, 148–149, 161, 163 and 166. The book's worked example is explicitly fictitious and simplified. Bibliographic record.
  2. RASB. Payment Orders Risk-Exposure Study (2025–2026) — public research description of controlled synthetic observations, three user experiences, and evidence limitations.
  3. BCBS. Operational risk and resilience, ORR10 — consolidated sound-management principles.
  4. BCBS. Principles for effective risk data aggregation and risk reporting (BCBS 239).
  5. EBA. Risk Assessment Report — December 2025 — CRR3 Business Indicator Component operational-risk capital context.
  6. RASB. The Method — public high-level conceptual description; differences between its payment-processing example and the published bank value-band implementation are tracked in the editorial approval log.

Explore Methodology & Empirical Research

Compare these definitions with the mathematical specifications on The Method or examine transaction-level operational risk quantification in the Payment Orders Study.

Discuss Academic Replication & Institutional Pilots →