Wall-to-Wall Physical Inventory: Cutoff, Tag Control and Variances

The count finished at 2 a.m., the floor team went home, and by Monday finance had a $61,000 shrink figure that nobody believed. Two trailers received after the freeze were counted in staging and then received again. Forty tags never came back. A pallet of customer-owned stock in the bulk aisle was counted as company stock. None of these is a counting mistake; each is a control that was missing before the first tag went out.

A wall-to-wall physical inventory holds up when three things are fixed in advance: one cutoff minute that freezes the book, a closed loop of numbered tags or WMS count tasks, and variance thresholds that decide who recounts and who approves the adjustment. With those in place, every line of the book-to-physical difference can be explained, and the journal posts on the morning after the count rather than a week later.

• What this guide covers: when a full count is the right tool, sizing teams and hours, freeze rules for receiving, shipping and transfers, tag control, blind counts and recounts, variance approval and the posting journal.
• What it adds: a labour-hours formula, a worked US ($) and India (₹) variance ledger with the CARO 10% class test and the GST credit reversal, and a readiness checklist for count night.

When a full shutdown count beats cycle counting, and when it does not

A wall-to-wall count earns its cost when the perpetual record cannot be trusted as a whole: after a WMS cutover, a site acquisition, a 3PL transition, or when the auditor will not rely on cycle-count results. Where cycle counts already cover every location with documented accuracy, a full shutdown mostly buys reassurance at the price of one or two lost shipping days.

The US audit standard frames the choice. PCAOB AS 2510, paragraph .09, says that when quantities are determined solely by a physical count at one date, the auditor ordinarily needs to be present at that count. Paragraph .11 accepts that some companies run controls, including statistical sampling, reliable enough to make an annual count of every item unnecessary, provided the auditor is satisfied the method gives substantially the same result.

• System change: opening balances migrated into a new WMS or ERP, where a full count gives a clean starting record instead of inherited errors.
• Ownership change: a 3PL handover, site purchase or client exit, where both parties need a signed quantity at a fixed hour.
• Failed cycle programme: location accuracy below the agreed target for two or more quarters, or cycle counts that skip slow movers and bulk storage.
• Auditor reliance: the external auditor plans to observe a year-end count rather than test the perpetual records.

If the question is how to keep accuracy high between full counts, our guide to cycle counting in a warehouse that keeps shipping covers that. The wave versus waveless release controls page deals with accuracy controls tied to order release. This article covers the single event: a shutdown count across every location, with the book frozen at a stated minute.

Sizing count teams, zones and hours from the location file

Plan count labour from the WMS location file, not from last year's headcount. Occupied locations, SKU-lot lines per location and the number of loose units drive the hours, and the usual planning error is to staff for pallets and then discover that each-pick shelving holds most of the lines.

Count labour hours per zone (two-person team = counter + recorder)
H_z   = (L_z × t_loc + N_z × t_line + E_z × t_each) / 3600 × (1 + r_z)
Teams = Σ H_z / (W × u)

L_z    = occupied locations in zone z (WMS location file)
N_z    = SKU-lot lines held in those locations
E_z    = loose units counted by hand (not full sealed cases)
t_loc  = seconds to travel to and identify a location
t_line = seconds to count and record one line
t_each = seconds per loose unit
r_z    = expected recount share of lines
W      = count window in hours
u      = productive share of the window

Illustrative each-pick shelving zone (time your own site):
L = 9,600   N = 12,800   E = 41,000
t_loc = 10  t_line = 25  t_each = 0.8   r = 0.06
H = (96,000 + 320,000 + 32,800) / 3600 × 1.06 = 132.1 team-hours
W = 8, u = 0.75  →  132.1 / 6.0 = 22.0  →  staff 23 teams

Run the formula separately for bulk pallet reserve, carton flow, each-pick shelving, cages and the yard, because the time per line differs by a factor of three or more. High-bay reserve needs a lift operator with every team, so it is often counted first, in daylight, under an early freeze on reserve replenishment. Add a control team of two or three experienced counters who do nothing but recounts.

Freezing the book: receiving, shipping and transfer cutoff rules

The freeze is a minute, not a shift. Every document posted before it belongs to the book; every movement after it waits. Goods that are physically present but on the wrong side of the line are either kept out of the count area or counted on tags that cite the document proving which side they sit on.

• Receiving cutoff: record the last goods receipt number posted. Trailers arriving later stay sealed, or are unloaded into a taped lane marked "after cutoff, do not count" with no receipt posted.
• Shipping cutoff: record the last ship-confirmed bill of lading or lorry receipt. Ship-confirmed goods still on the dock are sealed and excluded; staged goods without confirmation remain stock and are counted.
• Transfers and returns: inter-site shipments in transit are listed by document; customer returns stay in the returns cage and are counted on their own tags.
• Owned versus held stock: consigned or customer-owned goods are counted on tags carrying an owner code, so they never enter the company's variance.
• System hold: the WMS places every count zone in a hold status that rejects putaway, picking, replenishment and transfer transactions until the coordinator releases it.

Most cutoff failures happen at the dock, where daytime habits run into night rules. The trace below follows one illustrative failure from the freeze to the Monday close, and the control that would have stopped it.

1. 18:30, freeze declared: the coordinator logs last receipt GRN R-448210 and last ship confirmation BOL 77-30915; WMS zones move to COUNT-HOLD.
2. 19:12, late trailer: a carrier delivers 26 pallets on ASN 9902-118. The dock clerk unloads to staging lane D4 out of habit; no receipt posts and no tape goes down.
3. 21:40, staging counted: a team writes tags 006401 to 006426 for lane D4 as on-hand stock, because nothing marks the pallets as post-cutoff.
4. Monday 07:05, receipt posted: receiving posts ASN 9902-118 with transaction RCPT-ASN. The count already created a gain of 26 pallets, so the book now holds them twice, an illustrative overstatement of $48,100.
5. Control that stops it: post-cutoff goods taped and labelled at the dock, plus a rule that any tag in a staging lane must cite a pre-cutoff receipt number or be voided, and a review of every positive variance in staging before posting.

Pre-count readiness: consolidation, labels and open tasks

Most count-night recounts trace back to the week before: partial pallets spread across three locations, loose cartons without labels, and locations that exist on the floor but not in the WMS. A readiness walk five to seven days before the count, with a short punch list, removes more recounts than adding counters.

• Consolidate partials: merge part pallets of the same SKU and lot so one location holds one line where possible.
• Label everything: every location, LPN and loose carton carries a readable label; unlabelled cartons go to a problem cage before the freeze.
• Close open tasks: confirm or cancel pending putaways, pick shorts and replenishments so no unit is in a system state that does not match the floor.
• Fix the unit of measure: state on the tag or RF screen whether the count is in eaches, inner packs or cases, and verify case-pack quantities for the top-value SKUs.
• Separate status stock: damaged, quarantined and on-hold goods are counted on their own tags so their write-down follows the right approval path.

Tag control: issuing, voiding and reconciling every count tag

Every tag number printed must end the night in one of three states: used, voided with a reason, or returned unused. One unaccounted tag means nobody can prove the count is complete, or that a tag was not added after the fact.

Paper counts use pre-numbered two-part tags: the stub stays on the pallet or shelf face, and the upper part goes to the recorder. Tags are issued in logged blocks to each team. RF counts replace paper with WMS count tasks that carry unique IDs, and the same discipline applies: no team may create an unplanned count task without the zone lead's approval, which the system records.

• Issue log: tag range, team, zone and time issued, for example tags 006001 to 006500 to team 14 in zone D at 18:20.
• Void rule: a voided tag is kept, marked "void" with a reason and initials, never thrown away.
• Return check: unused tags come back with the team before anyone leaves the floor; the coordinator counts them against the issue log.
• Balance test: printed tags equal used plus voided plus returned. In an illustrative run of 19,500 tags, 17,842 used, 211 voided and 1,447 returned balance exactly.

Tag control in a wall-to-wall physical inventory reconciles every printed tag number in the WMS count log as used, voided or returned before the book-to-physical journal posts. PCAOB AS 2510 paragraph .09 expects the auditor to attend a single-date count, and one missing tag in a 19,500-tag run should hold the close.

Blind first counts, recount triggers and auditor test counts

The first count is blind: neither the paper tag nor the RF screen shows the expected quantity. Recounts follow a fixed rule on value and percentage, a different team performs them, and the external auditor's test counts are recorded by tag number so they can be traced to the final listing.

A practical recount rule is to recount any line where the absolute variance times unit cost is at least $250 or ₹20,000, or where the variance is at least 2% of book for high-value items and 5% for the rest. If the recount agrees with neither the first count nor the book, the control team performs a third count, and that count stands. The thresholds are illustrative; set them so recounts stay near 3% to 8% of lines.

Auditor test counts run both ways: from the floor to the tag listing, which tests completeness, and from the listing to the floor, which tests existence. The auditor records tag numbers and quantities and later compares them with the final count file. Any change to a tag after the auditor has recorded it needs a written reason, which is why tag edits belong to the zone lead and not the counter.

Variance approval bands and the book-to-physical journal

Approve variances by value band and inventory class, not by line count, and post one journal per class against the snapshot taken at the freeze minute. Report gross gains and gross losses beside the net, because a small net figure can hide large offsetting errors such as mis-slots between adjacent locations.

• Band one, small: net class variance under $1,000 or ₹80,000 is approved by the inventory control manager.
• Band two, material: $1,000 to $25,000, or ₹80,000 to ₹20 lakh, needs the site GM and the finance controller.
• Band three, escalated: anything larger, or any class off by more than 2% of book, goes to the CFO with a root-cause note.

The bands above are illustrative and should match the company's delegation of authority. The journal itself is plain: debit an inventory shrink or loss account and credit inventory for net losses, the reverse for net gains, at the cost basis the ledger already uses. Ind AS 2, paragraph 34, requires all losses of inventories to be recognised as an expense in the period the loss occurs, which rules out deferring an unexplained shortage to the next quarter.

India's CARO 2020 paragraph 3(ii)(a) asks the statutory auditor whether physical verification of inventory was adequate and whether discrepancies of 10% or more in the aggregate for each class were properly dealt with in the books, so the count file should report variance by class, not only by SKU.

Reporting duties after the count in the US and in India

One count serves different reporting duties in each market. In the US the focus is the auditor's observation under AS 2510 and, for SEC registrants, management's internal-control assessment under Sarbanes-Oxley section 404; in India, CARO 2020 adds a class-level discrepancy test and GST law requires reversing credit on goods written off.

Section 404, codified at 15 U.S.C. 7262, requires the annual report to include management's assessment of internal control over financial reporting, with auditor attestation for most larger filers. A count that reveals a $48,100 cutoff error is a control finding, not only an adjustment. If the count falls before year-end, AS 2510 paragraph .12 explains why the auditor will also test the movements between count date and balance sheet date.

• CARO class test: in the worked example, packing material is short by 11.17% of its book value, so the auditor must report whether that discrepancy was properly dealt with in the books.
• Bank stock statements: CARO 2020 paragraph 3(ii)(b) asks whether quarterly returns to banks agree with the books where working capital limits exceed ₹5 crore, so a large count adjustment should be explained to the lender before the next statement.
• GST credit: Section 17(5)(h) of the CGST Act blocks input tax credit on goods lost, stolen, destroyed or written off, so credit taken on the shortage is reversed.
• Expense timing: under both ASC 330 and Ind AS 2, the loss belongs to the period of the count, not to the period when someone finishes investigating it.

A GST registrant in India that writes off a count shortage must reverse the input tax credit on those goods, because CGST Act Section 17(5)(h) blocks credit on goods written off; in the worked example, ₹19.40 lakh of gross shortages at 18% produce a ₹3,49,200 reversal.

3PL sites: client-owned stock, observers and shrink allowances

At a 3PL site the stock usually belongs to the client, so the count result is evidence in a commercial relationship, not only an accounting entry. Agree the cutoff, the observer's rights and the shrink allowance in writing before count night, or the variance becomes a dispute.

• Owner codes on tags: each tag carries the client code, and multi-client aisles are counted client by client so a pallet cannot drift between owners.
• Observer access: the client or its auditor may watch and test count, but corrections go through the zone lead so the tag record stays intact.
• Shrink allowance: the contract states the tolerated loss, commonly as a percentage of throughput or stock value, and what the 3PL pays above it.
• Signed result: both parties sign the class totals and the tag balance before adjustments post to the client's inventory feed.

Report the count to each client in its own units and value basis, then post the adjustment through the same inventory feed the client uses daily, so its ERP and the 3PL's WMS land on the same quantity. Our page on inventory accuracy measures that do not hide location and status errors explains how to report the result without letting net figures mask location errors.

Count-night readiness checklist and questions for a WMS vendor

Run the checklist at seven days out, at one day out and at the freeze minute. Anything still open at the freeze minute either moves the count date or is written into the cutoff memo with an owner and an expected effect on the variance.

When evaluating a WMS for count support, ask the vendor to demonstrate a zone hold that rejects every movement type, blind count tasks, recount rules by value and percentage, owner codes on count tasks, and a journal that will not post until the task balance test passes. A demo on a test site with five zones and two clients shows more than a slide deck. For more site-level planning, see the Logistics and Supply Chain hub.

If you are preparing a count after a system change or a 3PL transition and want the cutoff, tag and journal controls designed alongside the WMS configuration, you can send us a short request for proposal with your location count, client mix and count window. We will reply with a count plan and the WMS settings it depends on.