Charta is on Google Play: practise on the go. Get the app
Charta

Explain how operational loss data should best be gathered and loss incidents reported, including what supervisors expect.

FRM® Part II · Operational Risk and Resilience · Risk Measurement and Assessment · FRM2.ORR.4.a

The idea

Operational loss data are the factual base for measurement, so how they are gathered matters as much as the models built on them. Good practice records every event in one firm-wide taxonomy, with event type, business line, root cause, and three separate dates: occurrence, discovery and accounting. Gross loss, recoveries and net loss are stored separately. Losses from one root cause are grouped into a single event before any threshold is applied. Near misses and non-accounted impacts are captured because they warn of control weaknesses. Supervisors expect completeness, accuracy, reconciliation to the general ledger, independent validation and timely escalation of significant events to senior management. The usual failure is under-reporting, caused by a blame culture or by tests applied to fragments of a larger event.

Formulas

Net operational loss after recoveries

net loss = gross loss − recoveries

gross loss
the direct write-off plus external costs such as legal fees, before recoveries
recoveries
all amounts recovered, from insurers and from third parties such as counterparties or customers

Gross, recoveries and net are stored or reported separately. Counting only the insurance recovery, or reporting net alone, are the usual errors.

A practice question

From the question bank, with the reason each wrong option is wrong.

In operational loss data collection, which description of a near miss is correct?

  1. An event that could have caused a loss but did not, recorded because it exposes a control weakness. - correct
  2. An event whose loss was fully recovered through insurance, which is therefore removed from the loss database.Why not: Confuses a near miss with a recovered loss; recoveries are recorded against the gross loss, they do not turn the event into a near miss.
  3. A loss that falls just below the firm's collection threshold, which is therefore ignored.Why not: Treats a near miss as a small loss under the threshold; a near miss is about the absence of loss, not its size.
  4. A loss that was discovered in a later period than the one in which it occurred.Why not: Confuses a near miss with a late-discovered loss; that is a difference between occurrence and discovery dates.

A near miss is an incident where controls or luck prevented a loss, or the loss was nil. Firms record them because they reveal the same causes as real losses, at no cost, and so give early warning.

A harder one

An internal fraud began in June 2023, was discovered in March 2025 and was booked as a write-off in the second quarter of 2025. Which date should the loss database normally treat as the date of occurrence?

  1. March 2025, because an event cannot be known to exist before it is discovered and confirmed by the investigation team.Why not: Uses the discovery date; discovery is a separate field and using it would place the loss in the wrong period for trend analysis.
  2. Second quarter 2025, because the loss only exists once it is recognised in the accounts.Why not: Uses the accounting date; accounting recognition lags the event and is also recorded separately.
  3. June 2023, the date the event started; discovery and accounting dates are stored separately. - correct
  4. The midpoint between June 2023 and March 2025, since the fraud was active throughout.Why not: Invents an averaged date; a single start date is recorded and the duration is captured through the other date fields.

Good practice records three dates: occurrence (or start), discovery and accounting. Using occurrence for analysis keeps the loss in the period when the control failure took place.

The hardest tier

One software defect caused five separate mis-postings over 18 months, each recorded as a loss of USD 40,000 (total USD 200,000). The firm's data collection threshold is USD 50,000 per entry. What treatment is consistent with supervisory expectations?

  1. Group the five entries as one event with one root cause, total USD 200,000, above the threshold. - correct
  2. Exclude all five entries because each is below the USD 50,000 threshold.Why not: Applies the threshold entry by entry; supervisors expect related losses from one cause to be grouped before testing against the threshold.
  3. Record the five as independent events of USD 40,000 each, to keep the severity distribution granular and the frequency count high.Why not: Treats events sharing one root cause as independent, which understates severity and overstates frequency.
  4. Record only the first entry of USD 40,000 and treat later ones as repeat reporting.Why not: Drops the later losses as duplicates; they are real, separate postings that belong in the grouped event total.

Losses linked by a common root cause are grouped into one event (here 5 x 40,000 = 200,000), dated at the first occurrence. Threshold tests and severity modelling should use the grouped amount.

In the app: 6 practice questions on this outcome, spaced review that brings it back before you forget it.

Start free →

More in Risk Measurement and Assessment