The Evidence That Contradicts Itself
Auditing a warehouse, you sample 25 pick records against inventory system entries. You find 3 discrepancies where the physical count is lower than the system count. The warehouse supervisor explains that 'stock is adjusted weekly, so small variances are normal.' You check the adjustment log: it shows 40 manual adjustments in the last month, all under 2% variance, all approved by the same supervisor. The procedure says adjustments over 1% require a second approver. All 40 are recorded as 'under 1%' but your own recalculation of 5 of them shows variances of 1.4%–1.9%. When you raise this, the supervisor says the percentages are 'calculated differently' and offers to show you the formula 'later.' Your time on site is nearly up.
1. What is the finding, and how strong is your evidence given the time constraint?
Read the worked answer
The finding is a control failure in inventory adjustment: (a) adjustments appear to be systematically understated to fall below the 1% second-approver threshold, (b) one person both performs and approves all adjustments, and (c) your independent recalculation contradicts the recorded percentages on 5 of 5 rechecked. That pattern - not a single error - is what makes it significant: 40/40 under threshold, single approver, and recorded values that don't reconcile.
On evidence strength: I have a defensible finding on the recalculation (5 of 5 rechecked disagree) and on the single-approver fact (log shows one approver). What I do not yet have is proof of intent or the full population - I rechecked 5 of 40. So I state the finding at the level the evidence supports: 'recorded variance percentages do not reconcile with recalculated values on 5 of 5 sampled; all 40 adjustments were approved by the same individual contrary to the two-approver requirement for >1%.' I do not assert fraud. I do flag it as a potential integrity issue requiring escalation, because understated variances could mask shrinkage.
I would not accept the 'calculated differently' explanation without seeing the formula now; an unavailable formula is an unverified claim, recorded as such.
2. You have limited time left. What do you do before leaving site, and what goes in the report?
Read the worked answer
Before leaving: (1) preserve evidence - copy or photograph the adjustment log and the 5 recalculations, record record IDs; (2) ask once, in writing or with a witness, for the calculation formula and note the response; (3) sample a few more adjustments if minutes allow to strengthen or weaken the pattern; (4) verbally brief the supervisor's manager that a significant finding will be reported and that evidence has been retained, so there are no surprises and no opportunity to 'lose' the log.
In the report: state the objective evidence (counts, record IDs, the 5 recalculations and results), the requirement (two-approver rule >1%), and the gap. Note the unresolved item (formula not provided) as an open point, not as a conclusion. Recommend escalation given the possible integrity dimension. Avoid overstating beyond the sample; recommend a full-population review as the follow-up. The discipline is: report what the evidence shows, flag what it might mean, and don't let the clock turn a serious pattern into a vague note.