Operational risk is the possibility of loss from failures of people, processes or systems, or external events. An order-error ledger separates unintended exposure from actual loss.
Operational risk
Operational risk is the possibility of loss when people, processes or systems fail, or an external event disrupts operations. For a quant desk, a correct trading signal is not enough: an order can be duplicated, a position omitted from a risk report, or a software release can send a valid calculation to the wrong account.
In the Basel Committee's banking definition, operational risk includes legal risk and excludes strategic and reputational risk. These are boundaries of that regulatory definition: an operational incident can still damage a firm's reputation. See Basel Framework OPE10.1.
A trade can be wrong before it loses money
Consider a fictional cash purchase. The desk intends to buy 100 units at USD 12 per unit, but an order-entry error submits 1,000 units. Assume every unit fills at USD 12, with no fees, bid–ask spread, financing cost or price impact. There are no other positions or cash flows.
At execution
Intended trade
Actual trade
Extra amount caused by the error
Units bought
100
1,000
900
Cash change, USD
−1,200
−12,000
−10,800
Asset value acquired, USD
+1,200
+12,000
+10,800
Immediate net mark effect, USD
0
0
0
The error creates 900 units of unintended exposure, worth USD 10,800 at the fill price. That cash has bought an asset of equal value. Calling the entire USD 10,800 an immediate loss confuses the size of the position with its profit or loss.
Now stipulate a later market price of USD 11 per unit, before the desk corrects the position. The intended holding would have lost USD 100; the actual holding loses USD 1,000. The incremental mark loss caused by carrying the extra units is USD 900:
900 extra units(1,000−100)×−1 USD per unit(11−12)=−900 USD.
The extra asset is now worth USD 9,900 against its USD 10,800 purchase outlay. These are mark losses; a sale at the assumed mark would realise them before costs.
For the extra 900 unintended units bought at USD 12, cash decreases by USD 10,800 and asset value increases by USD 10,800, so the immediate net mark effect is zero. At USD 11 the asset is worth USD 9,900, leaving an incremental mark loss of USD 900.
Only the unintended position is shown. Synthetic cash-and-asset ledger in USD, before fees, financing, spreads or correction costs.
The operational failure created the unintended position; the subsequent market risk determined its price change. These causal channels can coexist. The economic comparison above does not, by itself, determine how a particular institution must classify a loss in regulatory reporting.
If the price had instead risen to USD 13, the extra units would show a USD 900 mark gain. A favourable outcome would not make the order-entry process sound. Operational risk concerns failures that expose the business to loss; it is not necessary for every individual mistake to produce a negative realised outcome.
Follow the failure through the workflow
“Human error” identifies too little to repair the process. In this example, useful questions include whether units were clearly displayed, whether an unusual order needed approval, whether fills were reconciled to the approved instruction, and who could stop further orders.
Point in the workflow
A control matched to this example
What it cannot establish alone
Before submission
Check quantity and monetary size against the approved instruction; require independent approval for exceptions.
A limit can be set incorrectly or overridden.
After execution
Reconcile the instruction, execution report and booked position; assign an owner to unresolved differences.
A correct trade record does not prove the original instruction was authorised.
During response
Stop further erroneous submissions, establish the actual exposure and authorise the correction.
Stopping new orders does not close positions already filled.
After the incident
Preserve the sequence of events, identify the failed control and test the repair.
A quiet period after a patch does not prove that the failure cannot recur.
These are illustrative controls, not a prescribed checklist for every desk. The Basel Committee's 2021 sound-management principles, especially Principle 9, discuss approval authorities, separation of conflicting duties, transaction reconciliation and technology controls. A second person checking the same misleading screen may provide less independence than the procedure suggests.
The scope extends beyond order entry. Fraud, a failed settlement process, a cyber incident, a data-centre outage or a critical service-provider failure can all create operational exposure. Model risk can overlap: a sound formula with an incorrectly implemented unit conversion is both a model implementation problem and a failure in the operating process. The useful response follows the mechanism rather than stopping at a label.
Count events, measure consequences, examine the missing data
A loss record should distinguish the incident, its cause, the exposure created, the loss measured and any recovery. Otherwise a position's gross purchase amount can be mistaken for its loss, or a recovery can obscure the scale of the original event. A near miss can reveal a failed control even when it contributes no realised loss to a loss total.
A common statistical starting point separates frequency from severity. If N is the number of loss events in a chosen year and Xi is the nonnegative loss amount for event i, aggregate annual loss is
S=i=1∑NXi,
with S=0 when N=0. Under the additional assumptions that severities are independent and identically distributed, independent of the event count and have finite mean, with finite E[N],
E[S]=E[N]E[X1].
This identity is a modelling tool, not evidence that event counts and severities really are independent. Shared outages, changing systems and reporting thresholds can undermine a simple fit. McNeil, Frey and Embrechts emphasise the limited, heterogeneous and changing data available for operational-loss modelling. A short history without a major event is weak evidence about rare, severe failures.
The Frequency, Severity and Aggregate Losses lesson develops compound distributions and their moments under explicit assumptions. The full lesson requires membership; its public page provides a preview.
Do not equate an internal loss model with today's regulatory capital calculation. The current Basel Framework OPE25 sets out a standardised operational-risk capital approach. Older texts also explain the historical Basel II approaches. The applicable legal requirements, implementation dates and national discretions must be checked for the institution's jurisdiction.
Sources
Basel Committee on Banking Supervision. Basel Framework OPE10: Definitions and application, §10.1 and its legal-risk footnote. Version effective 1 January 2023, updated 5 July 2024; checked 26 October 2026.
Basel Committee on Banking Supervision. Basel Framework OPE25: Standardised approach, §§25.1–25.2 and the historical-AMA reference in §25.10. Version effective 1 January 2023, updated 5 July 2024; checked 26 October 2026.
McNeil, A. J., Frey, R. and Embrechts, P. (2015). Quantitative Risk Management: Concepts, Techniques and Tools, revised edition, Princeton University Press, Chapter 13, §13.1, pp. 503–505, and §13.2.2: operational-risk scope, data limitations and compound loss modelling.
