Audit and Control Foundations and Responsibilities · Lesson 2

From objective to risk to control

5 min read

Lesson objectives

  • Write a causal objective–risk–control chain for a transaction and distinguish a control objective from a control activity.
Reading tools

Before proposing a control, ask what objective it protects. For customer refunds, an objective might be to return the correct amount to the correct customer for an eligible return within the promised service period. Errors could arise from a mistyped account, a clerk using access to create an ineligible refund, or a wrong calculation after a promotion. Saying there is a risk in returns is too broad to guide a test. Write an event that could be observed: a refund may go to an account that does not belong to the customer because changing the destination needs no independent check, producing loss and a complaint. This statement gives a possible cause, event, and effect. Treat the cause as a hypothesis until records support it; the risk statement helps decide what evidence to seek.

Distinguish the control objective from its activity. The objective is the condition to achieve with reasonable assurance: refund money should not reach an unauthorized account. An activity might block changes after return approval, require a second approval of a documented change, or flag changes for review. Approval by itself is not a testable specification. Who approves, when, against what criterion, in which record, and what happens on rejection? A precise rule might say: if the refund account differs from the original payment method, the system pauses release until a supervisor other than the request creator checks customer identity and the reason for the exception; the decision and time are retained. That rule can prevent one failure but may delay a legitimate refund, so the exception route must be clear to the customer and staff.

Consider a hypothetical week of 40 returns. Thirty-six went to the original payment method; four went to another account. Of those four, three have supervisory approval and an identity-check record. One has only the note customer asked for it. One divided by 40 does not establish a 2.5 percent fraud rate. It establishes an unsupported exception in the displayed week, without proving intent or a broader frequency. A sound next step is to isolate the transaction, seek the original documents and contact the customer through a trusted channel, then investigate why release was possible without the record. If the system accepts any free-text note instead of the required proof, the defect is in the design. If a rule normally blocks release but a privileged user bypassed it without review, the design might be suitable while operation or override access is weak. The distinction changes both the recommendation and the next test.

Build a five-column chain: objective, risk event, possible cause, control activity, and evidence of operation. In the refund case, the objective is paying the right recipient; the event is transfer to an unauthorized account; the possible cause is an unverified destination change; the activity is a payment pause and independent review; and evidence includes the request, original payment destination, supervisor decision, and release time. If no decision record appears, do not immediately conclude that no review occurred; it may reside in another system. Find the relevant log and state any access limit. Likewise, three well-recorded exceptions do not show that all exceptions worked. Define the period, population, and selection before making an operating-effectiveness statement. This chain yields an audit question: does the design prevent unjustified recipient changes, and does management follow up the exceptions it permits?

A measure called unsupported refunds sounds simple, but its denominator needs definition. With 40 returns in a week and four transfers to alternative accounts, the alternative-route frequency among all requests is 4 divided by 40, or 10 percent. The one unsupported request is one out of those four exceptions, or 25 percent of the exception category. These rates answer different questions. The first describes how often the alternative route was used in the displayed set; the second describes missing evidence within that route. Before comparing weeks, ensure one period counts requests and the other does not count partial transfers repeatedly. Even with a stable denominator, a missing record alone is not a confirmed loss. The measure guides further testing and later monitoring if calculated consistently.

Lesson summary

Start with the process objective and a causal risk event, then design a control for that failure route and identify operating evidence.