At some point every controller with a large, well-controlled inventory asks the same question: do we really have to count every item, every year — or can a statistically designed sample carry the audit? The instinct says sampling; the worry says the audit firm will reject it in February. The answer sits in the auditing standards themselves, and it is more favorable than most finance teams expect — provided the plan is designed for the referee who will judge it.
The short answer: the standards say auditors may accept sampling, not must
Yes — auditors may accept statistical sampling for inventory. PCAOB AS 2510.11 and AU-C 501 permit it when the sampling plan is statistically valid, properly applied, and produces reasonable results. The auditor decides, so the plan's design — not the concept — determines acceptance. Involve your auditor before mobilization.
That distinction — may, not must — is the single most useful thing to understand before proposing a sampled count. The concept has been blessed for over fifty years. What gets rejected is not sampling; it is sampling plans that fail the conditions the standards tell the auditor to test. The rest of this article walks the chain of authority link by link, then maps those conditions to count design.
Link 1: PCAOB AS 2510 blesses statistical sampling
Start with the strictest rulebook — the standard governing public-company audits. AS 2510, Auditing Inventories (paragraph .11) says, verbatim:
"In recent years, some companies have developed inventory controls or methods of determining inventories, including statistical sampling, which are highly effective in determining inventory quantities and which are sufficiently reliable to make unnecessary an annual physical count of each item of inventory... If statistical sampling methods are used by the client in the taking of the physical inventory, the auditor must be satisfied that the sampling plan is reasonable and statistically valid, that it has been properly applied, and that the results are reasonable in the circumstances."
This language is not new. It descends unchanged from AU Section 331 in SAS No. 1 (1972), which itself codified even older practice. Companies have been running statistically sampled inventories with auditor acceptance since before most of today's auditors were born. For a side-by-side of how the public-company and private-company standards relate, see our AS 2510 vs AU-C 501 standards comparison.
Link 2: AU-C 501 carries the same permission into private-company audits
Private companies are audited under AICPA standards, and AU-C 501 — which mirrors the international standard ISA 501 — governs inventory. It requires the auditor to attend physical counting and perform their own test counts when inventory is material to the financial statements. But it also grants two permissions that matter enormously to operations teams:
- Counting on a different date. AU-C 501 expressly permits the physical count to happen at a date other than the financial-statement date. The auditor then evaluates whether changes in inventory between the count date and period-end are properly recorded — the roll-forward. A summer baseline count supporting a December year-end is squarely within the framework, provided the roll-forward procedures are disciplined.
- Perpetual records and cycle counts. The standard contemplates perpetual inventory records verified through physical counts or other testing — the regulatory foundation under every well-run cycle count program.
Link 3: Extrapolation is the auditor's own method — AU-C 530.13
The most common objection to a sampled count is the estimate itself: "the auditors will never rely on an extrapolated number." The audit sampling standard says otherwise. AU-C 530 (paragraph .13) requires auditors to project the misstatements found in a sample to the population they drew it from. Extrapolation is not a vendor shortcut that auditors grudgingly tolerate — it is literally the methodology auditors themselves are obligated to apply in their own testing.
The mechanic is plain: the misstatement observed in the sample is projected across the untested population, and the projected total is compared to tolerable misstatement. When a count team hands the audit firm a stratified random sample with a stated confidence interval, it is speaking the auditor's native language — the same logic that governs fixed-asset audit sampling and the evaluation of types of audit evidence generally.
Link 4: The IRS accepts it too
The tax rulebook points the same direction as the audit rulebook. Treas. Reg. 1.471-2 permits book inventories verified by physical counts at reasonable intervals, and Rev. Proc. 2011-42 formally sanctions statistical sampling for federal tax purposes. Financial-reporting standards and tax authority agree: a properly designed sample is a recognized way to establish inventory quantities, end to end.
The conditions: what makes a sampling plan auditors will accept
The four links above settle the concept. Acceptance turns entirely on design — the specific things the standards direct the auditor to be satisfied about. Mapped to practice:
| Condition the auditor tests | What it means in practice | Why the auditor needs it |
|---|---|---|
| Statistically valid plan | Random selection, defined strata, and a stated confidence level and precision — written down before the first item is counted. | Without statistical validity there is nothing to project. This is the first thing AS 2510.11 tells the auditor to be satisfied about. |
| Complete population | Every item, location, and unit has a known, non-zero chance of selection. The roster is physically validated before fieldwork. | You cannot extrapolate over inventory the sample never had a chance to reach — an incomplete population invalidates the projection. |
| Estimated portion can’t hide a material error | The un-counted remainder is small enough that even a conservative error assumption stays below tolerable misstatement — with an expansion trigger if results deteriorate. | The auditor’s job is material misstatement. A sampled design must show the estimate cannot conceal one. |
| Solid per-item evidence | Every sampled unit is counted end-to-end with documented, time-stamped count records. | The projection is only as good as the counts underneath it. Weak per-item evidence poisons the whole sample. |
| Roll-forward for off-date counts | Movement between the count date and the financial-statement date is reconciled and supported by the cycle-count control framework. | AU-C 501 permits off-date counts only when intervening changes are properly recorded. |
| Auditor’s own involvement | The audit team is invited before mobilization — to vet the plan, observe counts, and select their own test items. | The standards make the auditor the referee. Involvement converts the count into evidence they helped scope. |
Read that table as a design checklist, not a list of objections. None of the six conditions disputes the concept of sampling — each one specifies what a defensible plan already contains.
The only real risk is surprise: the no-surprise playbook
In practice, the failure mode that sinks sampled counts is not statistics — it is surprise. Auditors accept sampling plans they were consulted on; they get prickly about estimates that show up in the trial balance unannounced. A sampled August count that the audit firm first hears about in February is "a number the auditors will interrogate." The same count, shared before mobilization, becomes "evidence the auditors helped scope."
- Share the draft sampling design early — population definition, strata, selection method, confidence and precision targets — before anyone counts anything.
- Agree the parameters with the audit team so confidence, precision, and stratification reflect their materiality, not just yours.
- Invite them to observe — a pilot site or the first count wave, where they can select their own test items.
- Document per-item evidence — time-stamped count records for every sampled unit, retained as workpapers.
- Reconcile and roll forward — deliver the projection with its confidence interval, the reconciliation to books, and the movement bridge to period-end.
This is the operating model CPCON runs on statistically sampled engagements: on multi-site programs we have executed whole-unit random-sampling designs — units selected randomly against a physically validated roster, counted item-by-item with time-stamped records, and projected with a stated confidence interval. When a deal, a lender, or an audit deadline is driving the clock, that design work is the core of our independent inventory verification services, and the same evidence discipline underpins the existence assertion under SOX 404.
When sampling is not the right call
Honesty about the boundaries is part of a defensible plan. Sampling is usually the wrong answer when:
- the population is high-variance and individually significant — a few items carry material value, so they should be counted 100% as key items, not sampled;
- perpetual records are weak or the book population itself is unreliable;
- there is no internal-control history for the auditor to lean on — including first-year audits with no baseline;
- prior counts surfaced problems the auditor has already flagged.
In those situations the honest recommendation is a full wall-to-wall count or expanded testing — build the control history first, and earn the sampled design in later periods.
Frequently asked questions
Will auditors accept statistical sampling for a physical inventory count?+
Yes — auditors may accept it. PCAOB AS 2510.11 and AU-C 501 permit sampling when the plan is statistically valid, properly applied, and yields reasonable results. Acceptance is the auditor’s judgment call, so plan design and early involvement matter most.
What does the auditor require before relying on a sampling plan?+
A statistically valid plan (random selection, defined strata, stated confidence and precision), a complete population, un-counted risk bounded below tolerable misstatement, documented per-item procedures, a roll-forward for off-date counts, and the auditor’s own observation and test counts.
Can we count inventory on a date other than the financial-statement date?+
Yes. AU-C 501 permits counting at a date other than the financial-statement date, provided you perform a reconciled roll-forward of inventory movement between the count date and period-end.
Is extrapolation from a sample reliable enough for auditors?+
Extrapolation is the auditor’s own mandated method — AU-C 530.13 requires auditors to project sample misstatements to the population. Projecting a valid sample is standard audit procedure, not a vendor shortcut.
Does statistical sampling satisfy the IRS as well as financial auditors?+
Yes. Treas. Reg. 1.471-2 permits book inventories verified by physical counts at reasonable intervals, and Rev. Proc. 2011-42 formally sanctions statistical sampling for federal tax purposes.
Can cycle counts replace an annual wall-to-wall count?+
They can. AU-C 501 contemplates perpetual records verified by cycle counts, and AS 2510 recognizes methods reliable enough to make an annual count of every item unnecessary — subject to strong internal controls and the auditor’s satisfaction.
The bottom line
Every rulebook that touches the question — PCAOB, AICPA, and the IRS — points the same direction: a statistically valid, properly applied sampling plan is an accepted way to establish inventory quantities. The standards make the auditor the referee, so the work is in the design and the timing: build the six conditions into the plan, and put it in front of the audit team before mobilization, not after. Do that, and "will our auditors accept it?" stops being a risk and becomes a scoping conversation.
