Can cycle counting replace the annual physical inventory?
Yes, if your perpetual records are well kept and your counting program is reliable enough to give results substantially the same as counting everything once a year. That test comes from the PCAOB’s AS 2510 for public-company audits. For private companies, the AICPA’s AU-C 501 says the effectiveness of your controls over changes in inventory decides whether counts at other dates are appropriate. Either way, the auditor still observes some counts, tests your records and checks the movement between the last count and year-end. Cycle counting does not remove the evidence an auditor needs. It spreads that evidence across the year.
The decision belongs to your auditor, not to your ERP vendor or a blog post. What you control is the evidence. This field note quotes the standards that govern the decision, translates them into the records an audit team will ask for, and lists the conditions that still push a company back to a wall-to-wall count. The terms used throughout:
- Physical inventory (also called a wall-to-wall count or annual physical inventory): the ASCM Supply Chain Dictionary defines it as “the determination of inventory quantity by actual count,” and notes that physical inventories “can be taken on a continuous, periodic, or annual basis.”
- Cycle counting: in the same dictionary, “an inventory accuracy audit technique in which inventory is counted on a cyclic schedule rather than once a year.”
- Perpetual inventory record: “a computer record or manual document on which each inventory transaction is posted so that a current record of the inventory is maintained.”
Those definitions are from the ASCM Supply Chain Dictionary, 18th edition (2024). The question in this note is when the second can stand in for the first in a financial statement audit.
What the auditing standards say
Both US auditing standards treat observation of inventory counts as the normal procedure, and both allow counts during the year when the records and controls support it. They differ in wording, not in substance.
Public companies: PCAOB AS 2510
AS 2510, Auditing Inventories governs audits under PCAOB standards. Its opening line sets the default: “Observation of inventories is a generally accepted auditing procedure.” An auditor who issues an opinion without it “has the burden of justifying the opinion expressed.”
Three paragraphs carry the cycle counting question:
- Paragraph .09 covers the traditional case. When quantities are determined “solely by means of a physical count” at or near the balance-sheet date, the auditor ordinarily needs to be present at the count.
- Paragraph .10 covers perpetual records with counts during the year: “When the well-kept perpetual inventory records are checked by the client periodically by comparisons with physical counts, the auditor’s observation procedures usually can be performed either during or after the end of the period under audit.”
- Paragraph .11 covers methods that replace the annual count of every item. It says some companies have methods, “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.” In that case “the independent auditor must satisfy himself that the client’s procedures or methods are sufficiently reliable to produce results substantially the same as those which would be obtained by a count of all items each year.” The auditor “must be present to observe such counts as he deems necessary.”
Paragraph .11 also sets the bar for statistical sampling: 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.” And paragraph .12 is the fallback. If the auditor is not satisfied, “tests of the accounting records alone will not be sufficient,” and the auditor will need to “make, or observe, some physical counts of the inventory and apply appropriate tests of intervening transactions.”
Private companies: AICPA AU-C 501
Audits of nonissuers, meaning companies whose audits are not required to follow PCAOB standards, follow the AICPA’s Statements on Auditing Standards. AU-C section 501, in the AICPA’s codified standards current as of August 2026, requires the auditor, when inventory is material, to obtain evidence about its existence and condition by “attending physical inventory counting, unless impracticable.” At the count the auditor evaluates management’s instructions, observes the count, inspects the inventory and performs test counts (paragraph .12).
The application guidance is where cycle counting fits:
- Paragraph .A21: “Management ordinarily establishes procedures under which inventory is physically counted at least once per year to serve as a basis for the preparation of the financial statements and, if applicable, to ascertain the reliability of the entity’s perpetual inventory system.”
- Paragraph .A32: counting “may be conducted at a date, or dates, other than the date of the financial statements,” whether management uses an annual count or “maintains a perpetual inventory system.” Then the key sentence: “the effectiveness of the design, implementation, and maintenance of controls over changes in inventory determines whether the conduct of physical inventory counting at a date (or dates) other than the date of the financial statements is appropriate for audit purposes.”
- Paragraph .A33: differences between the perpetual records and the physical quantities “may indicate that the controls over changes in inventory are not operating effectively.”
When counts happen at dates other than year-end, paragraph .13 adds a requirement: procedures to obtain evidence “about whether changes in inventory between the count date and the date of the financial statements are recorded properly.”
The tax rule in the background
Federal tax regulations point the same way. Under 26 CFR 1.471-2(d), where a taxpayer keeps book inventories under a sound accounting system, “the balances shown by such book inventories should be verified by physical inventories at reasonable intervals and adjusted to conform therewith.” The regulation does not say “annual” or “wall-to-wall.” Talk to your tax adviser about how your count program supports that sentence.
What the standards do not say
Several things that vendor pages imply are not in either standard:
- No accuracy threshold. Neither standard names a percentage at which cycle counts become acceptable. “Substantially the same” is a judgment the auditor makes from your evidence.
- No required count frequency. AU-C 501.A21 says management “ordinarily” counts “at least once per year.” A program that counts every item at least once a year appears to fit that sentence, but the reading is your auditor’s to make. Neither standard says how often A, B or C items must be counted.
- No automatic acceptance. A cycle counting program does not oblige the auditor to accept it. AS 2510.11 puts the burden on the auditor to “satisfy himself,” which in practice means on you to supply the evidence.
- No ban on keeping a full count. A company can cycle count and still take a year-end count, by choice or because a lender or owner wants one.
Read the standards as describing what the auditor must be satisfied of, not a checklist that approves your program once each box is ticked.
The standards side by side
| Question | PCAOB AS 2510 (public companies) | AICPA AU-C 501 (private companies) |
|---|---|---|
| Default procedure | Observation of inventories is “a generally accepted auditing procedure” | Attend physical inventory counting “unless impracticable” |
| When counts during the year work | Well-kept perpetual records “checked by the client periodically by comparisons with physical counts” | When controls over changes in inventory are effective in design, implementation and maintenance |
| Replacing the annual count of every item | Methods must produce “results substantially the same” as counting all items each year | Counts at “a date, or dates” other than year-end are allowed when controls support it |
| What the auditor still does | Observes counts as deemed necessary; tests the counting procedures | Evaluates instructions, observes, inspects, performs test counts |
| Year-end gap | “Appropriate tests of intervening transactions” when not otherwise satisfied | Procedures on changes between count date and year-end |
Where the PCAOB inventory project stands
AS 2510 has not changed, and as of the latest PCAOB agenda we could read (August 2026), no revision is scheduled. Plan around the current text.
The current text is old. The archived predecessor, AU section 331, carries an issue date of November 1972. That is why paragraph .11 speaks of methods developed “in recent years.” The PCAOB’s own inventory project page (updated November 18, 2024) says AS 2510 “was adopted as an interim standard in 2003,” and that its requirements “are some of the oldest among those currently codified within PCAOB standards.”
The project’s stated objective was to “consider changes to AS 2510, Auditing Inventories, to address developments in inventory management processes of companies and increasing use of technology tools by companies and by auditors.” The project page’s status line read: “The staff is analyzing relevant information and developing a proposal for the Board’s consideration.” Archived copies of the PCAOB’s standard-setting agenda from January 2025 through January 2026 listed the inventory project with a proposal anticipated in 2025.
The agenda then changed. By June 2026 the page noted that on March 31, 2026, the Board had issued a request for comment on its strategic priorities, and the interim agenda no longer listed the inventory project. The August 2026 copy notes a further request for comment on standard setting, issued June 23, 2026, and still lists no inventory project. The PCAOB’s live site blocks automated readers, so these checks used archived copies. Check the agenda page yourself before you rely on it.
For a controller, the conclusion is practical. The 1972 text, and its “substantially the same” test, is what your public-company auditor applies today. Private-company audits follow AU-C 501 as quoted above.
What “well-kept perpetual records” means in practice
Well-kept perpetual records are records in which every movement of stock is posted as it happens, to the right item, location and quantity, by someone allowed to post it. Neither standard lists the controls, so the list below is a practitioner’s translation, not quoted text.
- Every movement is a transaction. Receipts, put-aways, moves, picks, shipments, returns, scrap and production consumption each post to the record. Stock that moves without a transaction is the main source of record error.
- Transactions post at the time and place they happen. Scanning the location and item at the moment of the move keeps the record current. Paperwork keyed at the end of the shift opens a window in which the record is wrong.
- Cutoff is controlled. Receipts and shipments around any count, and around period end, are recorded in the right period. AU-C 501 lists “control over the movement of inventory between areas and the shipping and receipt of inventory before and after the cut-off date” among the matters that count instructions should address.
- Adjustments are approved and explained. No one posts an adjustment to stock they counted without a second person’s approval above a threshold. Each adjustment carries a reason code. Our note on inventory adjustment reason codes and approval rules covers the code set and thresholds.
- Negative balances are treated as errors. A record that allows stock to go below zero, and leaves it there, is not well kept.
- Access is limited. Only defined roles can change on-hand quantities, item units of measure or location records, and those changes are logged.
Research supports the link between counting discipline and record quality. In a study of nearly 370,000 inventory records from 37 stores of one retailer, DeHoratius and Raman (2008) found 65 percent inaccurate, and identified inventory auditing practices among the factors that reduce record inaccuracy. That was a retail study, so read it as evidence of the mechanism, not as a benchmark for a distribution center.
The evidence file your auditor will ask for
Build the evidence before the audit planning meeting, not after it. The table maps each phrase in the standards to the records that answer it and the system that produces them.
| What the standard asks | Evidence you hand over | Where it comes from |
|---|---|---|
| "Well-kept perpetual inventory records" (AS 2510.10) | Transaction controls: scanned receipts, moves and picks; approved adjustments; access roles | ERP and WMS transaction logs, user access report |
| Records "checked ... periodically by comparisons with physical counts" (AS 2510.10) | The count plan and 12 months of count history covering every item | Cycle count plan, count lines with system and counted quantities |
| "Results substantially the same" as counting all items each year (AS 2510.11) | Accuracy by location, class and value, month by month | Accuracy report built from count lines |
| Effectiveness of "the counting procedures used" (AS 2510.11); "count and recount procedures" (AU-C 501.A24) | Written count instructions: blind counts, recount rule, tolerances | Count procedure document, system settings |
| Differences "may indicate that the controls over changes in inventory are not operating effectively" (AU-C 501.A33) | Variance investigations with cause, fix and follow-up count | Variance log, adjustment approvals |
| Changes "between the count date and the date of the financial statements" (AU-C 501.13); "tests of intervening transactions" (AS 2510.12) | Cutoff controls and a roll-forward from count dates to year-end | Receiving and shipping cutoff reports, transaction history |
| Inventory "in the hands of public warehouses or other outside custodians" (AS 2510.14; AU-C 501.16) | Custodian confirmations and the custodian's count records | Third-party warehouse reports |
The count history, field by field
The count history is the core of the file. For each count line it should show:
- Item, location, lot or serial number where tracked, and owner.
- System quantity at the moment of the count, and counted quantity.
- Whether the count was blind (the counter did not see the system quantity).
- Counter ID, date and time.
- Recount quantity and recounter, where a recount happened.
- Variance in units and in value, and whether it was within tolerance.
- The adjustment posted, its reason code and approver, or a note that no adjustment was needed.
A count file that holds all seven lets an auditor reperform your accuracy calculation and trace a sample of counts to adjustments. A spreadsheet assembled from memory at year-end does not.
Who prepares which part
The file crosses three teams, and it is complete only when each knows its part:
- Inventory manager: the count procedure, the schedule, the count history and the variance log.
- Controller: the accuracy reports by value, the adjustment approvals, the roll-forward and the conversation with the audit team.
- ERP or WMS administrator: the transaction and access reports, the system settings for blind counts and tolerances, and the cutoff reports.
Name an owner for each item in the table above and agree a date by which it is ready for the audit planning meeting.
The count procedure, in writing
AU-C 501 says the auditor evaluates “management’s instructions and procedures for recording and controlling the results of the entity’s physical inventory counting.” For a cycle counting program, those instructions are a standing document, not a memo for one weekend. It should cover the count schedule and classes, blind counting, tolerances by class, the recount rule, the investigation steps before adjustment, approval thresholds, and how counts interact with picking and receiving in the same location. The steps before adjustment are covered in investigating a cycle count variance before you adjust.
Designing a count program that holds up
A count program holds up when every item is counted on a schedule that matches its value and risk, the counts happen every working day, and the auditor can see the schedule kept. The ASCM dictionary describes the same shape: “Most effective cycle counting systems require the counting of a certain number of items every workday, with each item counted at a prescribed frequency.”
Classes and frequencies
Most programs start from ABC classification. The ASCM dictionary describes the A group as usually “10 percent to 20 percent by number of items and 50 percent to 70 percent by projected dollar volume,” and states the principle that “effort and money can be saved through applying looser controls to the low-dollar-volume class items than to the high-dollar-volume class items.” Count A items most often, C items least, and make sure every item gets at least one count a year. An item never counted cannot support a claim that the results are “substantially the same” as counting everything.
Add counts triggered by events: a location that should be empty but is not, a pick from an empty bin, a negative balance, a large adjustment. Event counts are cheap, and they catch errors close to when they happened.
Count locations, not just items
Count by location as well as by item, including locations the system says are empty. Item-based counts confirm that the item is where the system says. Location-based counts also find stock the system does not know about. For a test of existence and completeness, an auditor will want both directions. AU-C 501.A27 describes test counts as “tracing items selected from management’s count records to the physical inventory and tracing items selected from the physical inventory to management’s count records.”
Example (fictional). A distributor has 6,000 active items. It counts 500 A items monthly (6,000 counts a year), 1,500 B items quarterly (6,000 counts) and 4,000 C items once (4,000 counts), for 16,000 counts a year. Over 250 counting days that is 64 counts a day. At 30 counts an hour, one counter needs a little over two hours a day, before recounts and investigations. You can size your own program in the cycle count planner.
Counting while the warehouse is running
A wall-to-wall count often means stopping receiving and shipping. A cycle count does not, so the program has to control movement around each count instead. AU-C 501.A24 lists “control over the movement of inventory between areas and the shipping and receipt of inventory before and after the cut-off date” among the things count instructions should address. For cycle counts, that cutoff happens at every location, every day.
Two designs work. Either the system blocks picks, put-aways and moves at a location while it is being counted, or it records the system quantity at the moment of each scan and lists the transactions that touched the location during the count, so the investigator can account for them. Whichever your system does, write it into the count procedure, because the auditor will ask how a count can be right while stock is moving.
Statistical sampling instead of counting everything
AS 2510.11 also allows a different design: estimating inventory from a statistical sample rather than counting every item at some point in the year. The bar is specific. 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.”
In practice that means a defined population (every item and location in scope), random selection rather than convenient locations, a documented method for projecting the sample’s errors to the whole population, and a pre-agreed response when the projected error is too large. Plan to involve a statistician, and agree the design with the audit team before the first count. A full-coverage cycle count program is simpler to explain and defend, because every item is actually counted.
Running both in the first year
The most direct evidence that cycle counting gives “results substantially the same” as a full count is to run both for one year. Cycle count all year, then take a final wall-to-wall count and compare it with the perpetual records the cycle counts maintained. If the year-end count finds only small, explained differences, the comparison answers the auditor’s main question with your own data. If it finds large differences, you have learned where the program is weak before relying on it alone. Discuss this transition plan with your auditor during planning, so both sides agree in advance what result would support dropping the full count the following year.
Schedule the auditor’s visits
The auditor will want to observe counts. AS 2510.11 says the auditor “must be present to observe such counts as he deems necessary.” Agree on the visits during audit planning, and pick counts that represent the program: a normal day, with the normal team and the normal instructions. A count staged for the visit tells the auditor less than an ordinary one.
How accurate is accurate enough
Neither AS 2510 nor AU-C 501 sets an accuracy percentage. The test is whether your method gives “results substantially the same” as a full count, and whether differences point to weak controls. Agree how accuracy will be measured with your auditor before the year starts, then report it the same way every month.
ASCM defines inventory accuracy as the state “when the on-hand quantity is within an allowed tolerance of the recorded balance,” usually “measured as the percent of items with inventory levels that fall within tolerance.” It adds that “target values usually are 95 percent to 99 percent, depending on the value of the item.” Those are operating targets from a professional body, not audit thresholds. They are still a reasonable starting point for the conversation.
Measure three ways
One accuracy number hides too much. Report at least three:
| Measure | What it shows | Why the auditor cares |
|---|---|---|
| Count accuracy by location or line | Share of counts within tolerance | Whether the records are reliable item by item |
| Net value variance | Gains minus losses, as a percent of counted value | The likely effect on the balance sheet |
| Gross value variance | Gains plus losses, as a percent of counted value | How much error the net figure is hiding |
Break each measure down by class, zone and month. A program with good overall accuracy and one zone that keeps failing has a control problem in that zone. AU-C 501.A33 says such differences “may indicate that the controls over changes in inventory are not operating effectively.”
Set tolerances by class, in writing
Tolerance is the difference you treat as accurate. ASCM ties it to “the value of the item,” so set it by class and write it into the count procedure before the year starts. Changing tolerances mid-year makes the accuracy trend impossible to compare. For example, a hypothetical starting set might look like this:
| Class | Hypothetical tolerance | Reasoning |
|---|---|---|
| A (high value) | Zero units | Any difference on a high-value item matters to the balance |
| B | 1 percent of quantity, or 1 unit | Small handling differences on mid-value items |
| C (low value, small parts) | 3 percent of quantity | Counted by weight or estimate; small errors carry little value |
Report accuracy both within tolerance and at zero tolerance for the first year, so the audit team can see what the tolerance hides.
Example (fictional). In October a manufacturer counts 1,320 locations. Of those, 1,268 are within tolerance: 96.1 percent count accuracy. The counted value is $1.92 million. Gains and losses net to $4,180 (0.22 percent), but together they total $11,900 (0.62 percent). The gross figure is almost three times the net. Most of the gross variance sits in one zone, the line-side locations for the assembly cell, where backflushed components and late production reports disagree. The controller shows this to the audit team along with the fix, rather than waiting for them to find it.
You can run the location and value measures from a count file in the inventory accuracy calculator.
What still triggers a full count
Some conditions undermine the evidence a cycle counting program provides. When one applies, your auditor may ask for a full count, a one-time count of the affected area, or more extensive testing. The list below links each condition to the standard it touches.
- Accuracy results below what was agreed, or a falling trend. The “substantially the same” test in AS 2510.11 is no longer met.
- Control failures in the perpetual records. Large unexplained adjustments, persistent negative balances, or transactions posted by people who should not post them. AU-C 501.A33 ties such differences to controls over changes in inventory.
- A system change during the year. A new ERP or WMS, or a re-slotting of the warehouse, breaks the continuity of the count history. Opening balances in a new system are themselves a count question, covered in moving on-hand inventory into a new WMS.
- A first-year audit or a change of auditor. AS 2510.13 deals with auditing periods in which the auditor “had not observed or made some physical counts of prior inventories,” and says the auditor may become satisfied through procedures such as “tests of prior transactions, reviews of the records of prior counts, and the application of gross profit tests.” A complete count history makes those procedures possible.
- Locations without perpetual records, such as a satellite site or a maintenance crib where stock is issued without transactions. Those locations have nothing for cycle counts to check against.
- A count program that skipped items, or counted the same easy locations repeatedly. Coverage gaps undermine the claim that results match a full count.
- The auditor cannot observe. AU-C 501.A35 notes that “the matter of general inconvenience to the auditor, however, is not sufficient to support a decision by the auditor that attendance is impracticable.” Plan access for count visits as carefully as you would for a year-end count.
Some companies keep a full count for business reasons even when the audit does not need one: a lender’s requirement, an insurance claim, a sale of the business. Check your agreements before you announce the end of the annual count.
Inventory at third-party warehouses
Stock held by a third-party warehouse or other custodian is audited differently. The auditor usually confirms it with the custodian, and may add other procedures when it is significant.
AS 2510.14 says that if inventories are “in the hands of public warehouses or other outside custodians, the auditor ordinarily would obtain direct confirmation in writing from the custodian.” When such inventory is a significant share of assets, the auditor applies one or more further procedures, including testing “the owner’s procedures for investigating the warehouseman and evaluating the warehouseman’s performance,” obtaining an independent accountant’s report on the warehouse’s controls, and observing “physical counts of the goods, if practicable and reasonable.” AU-C 501.16 gives private-company auditors a similar choice: request confirmation from the third party, or “perform inspection or other audit procedures appropriate in the circumstances.”
For a company that cycle counts its own buildings and stores stock at a third-party warehouse, that means two evidence files:
- Your own sites: the count program and evidence file described above.
- The custodian’s sites: the custodian’s count program and results, its adjustment reports, your reconciliation of its records to yours, and the confirmation process. If the warehouse sends inventory adjustment messages (X12 947, “Warehouse Inventory Adjustment Advice,” in ASC X12’s list), keep them and reconcile them monthly.
Ask the third-party warehouse early in the year what count records it can provide by client. The third-party logistics page covers owner-level stock records in more detail.
Closing the year without a wall-to-wall count
Without a year-end count, the year-end balance rests on the perpetual records, so the auditor tests the transactions between the last counts and the balance-sheet date. Prepare for that test the way you would prepare for a count weekend.
AU-C 501.13 requires procedures on “whether changes in inventory between the count date and the date of the financial statements are recorded properly.” AU-C 501.A34 lists the relevant matters: “Whether the perpetual inventory records are properly adjusted,” the “reliability of the entity’s perpetual inventory records,” and the “reasons for significant differences between the information obtained during the physical count and the perpetual inventory records.” Under AS 2510, the equivalent language is “appropriate tests of intervening transactions.”
What to prepare for year-end
- A cutoff procedure for the last days of the year. Record the last receiving and shipping documents processed before the balance-sheet date, and hold or clearly date anything that arrives or leaves around midnight. Copies of cutoff information help the auditor test it later; AU-C 501.A25 notes the auditor “may obtain copies of cutoff information, such as details of the movement of inventory.”
- Goods in transit and on the dock. List inbound receipts not yet put away, outbound shipments staged but not shipped, and transfers in transit between sites.
- A roll-forward by class. For a sample of items, show the last count, every transaction since, and the year-end balance. Auditors test this, so make it a standard report.
- Open adjustments and investigations. Close or document every variance still under investigation at year-end, with its value.
- Consigned and customer-owned stock. List stock you hold that you do not own, and stock you own held elsewhere, so neither ends up in the wrong balance sheet. AU-C 501.A24 names “inventory owned by a third party (for example, on consignment)” among the matters count instructions should address.
Example (fictional). A distributor switches from a December wall-to-wall count to cycle counting. In the planning meeting the audit team agrees to observe two count days, in May and October, and to test a roll-forward from the last count of 40 sampled items to December 31. The controller adds a December 31 cutoff log at both docks and a weekly report of open variances. In January the auditor selects items from the count history, traces them to the shelf and back, and tests the roll-forward. No warehouse shutdown is needed.
What to do before you talk to your auditor
Bring evidence, not a request. Six months of count history, an accuracy trend, a written procedure and a cutoff plan let the audit team assess the program on facts. A proposal to stop counting without them is easy to refuse.
A checklist for the controller and the inventory manager:
- Write the count procedure: classes, frequencies, blind counting, tolerances, recount rule, investigation steps, approval thresholds.
- Check coverage: every active item and every location gets at least one count a year. Size the workload in the cycle count planner.
- Fix the count file: each count line carries the seven fields listed above, kept in the system, not rebuilt in a spreadsheet.
- Report accuracy three ways (count, net value, gross value) by class, zone and month, starting now.
- Log every variance investigation with cause, fix and follow-up count.
- Review adjustment controls: reason codes, approval thresholds, and who can post.
- Plan cutoff and roll-forward reports for year-end, and test them at a quarter-end first.
- Inventory third-party sites: list them, their counts and their confirmations.
- Meet the audit team before the year begins, with the procedure and six months of results, and agree how accuracy will be measured and which counts they will observe.
If your systems cannot produce the count history, the accuracy breakdowns or the roll-forward without manual work, that is a records problem to solve before the audit conversation. It is the kind of work described on our inventory management software page. In our engagements we start by writing down the count procedure and the reports it needs, so the system is built to produce the evidence the audit will ask for.
