How often should a warehouse inventory audit be done?

Most operations run a continuous cycle count program and a full physical inventory once a year, often because their auditors or lenders require it. A data audit of the records behind the counts is worth doing annually, and immediately after a WMS go-live, a large write-off, or a sustained fall in count accuracy.

Can we audit inventory without shutting the warehouse down?

A data audit needs no shutdown at all; it works from exports. A count audit can run live if counts are blind, the system quantity is captured at count time, and locations with open tasks are excluded or recounted. A wall-to-wall count for financial purposes usually does need a cut-off and a freeze.

What does a warehouse inventory audit cost?

Counting services are generally priced on labor, so cost follows SKU count, location count and hours on site. A data audit is priced on scope. WMSAudit's fixed fee is $1,500 to $2,500, quoted before work starts, and doesn't change if the data turns out to be harder than expected.

Is an inventory accuracy audit the same thing?

Usually it means the same job with the emphasis on the measurement: is the accuracy figure you report true, and at what level? A good one recomputes accuracy from the raw count file at location level, in absolute terms, and then explains the misses rather than just reporting them.

Service guide · Inventory audit

Warehouse inventory audit:
counting the stock vs auditing the record

“Inventory audit” covers two very different jobs. One counts what is on the racks. The other works out why the system disagrees with it. Most warehouses buy the first when their problem is the second, and pay for another count next year.

Updated 17 September 2026
Written by WMSAudit
A Ravenspire LLC company
10 min read

Short answer

A warehouse inventory audit checks whether recorded inventory matches physical stock. There are two kinds. A count audit, meaning a wall-to-wall or sampled physical count, measures and corrects the gap on one day. A data audit examines receipts, item master, locations, adjustments and interfaces to find why the gap exists. The count fixes today's number. The data audit is what stops next quarter's variance.

Two kinds of inventory audit

Search for a warehouse inventory audit and you'll find two different industries using the same words. One sends a crew with scanners to count every location over a weekend. The other asks for your data and never sets foot on the floor. Both are called an inventory audit, and they answer different questions.

The count audit: A wall-to-wall physical inventory, or a statistically sampled count, compared against the system. It answers how far off are we today? Its output is a variance list and, usually, a set of adjustments that bring the book back into line with the floor. It is what your financial auditors care about at year end, and it is the right tool when you need a trustworthy opening balance.
The data audit: An examination of the records and transactions that produce the on-hand balance: receipts, the item master, the location master, moves, picks, counts, adjustments and interface messages. It answers why are we off, and will it happen again? Its output is a list of mechanisms, each with the items affected, the date it started, the owner who can fix it and what it costs a year.

The difference matters because a count corrects the balance and leaves the mechanism running. If the variance was caused by a wrong case pack, a duplicate-receipt path or keyed putaway confirmations, the count adjusts it away and the same process rebuilds it, often by the same amount, on the same items, within weeks.

Count audit vs data audit

Comparison of a physical count audit and a WMS data audit
	Count audit	Data audit
Question it answers	How far off are we today?	Why are we off, and will it recur?
Works from	The racks	Exports: receipts, item master, locations, transactions, adjustments, interface logs
Disruption	Labor on the floor; often a freeze or cut-off	None on the floor; a data pull
Output	Variance list and adjustments	Mechanisms, affected items, onset date, owner, annual cost
Effect on the balance	Corrects it on the day	None directly; stops the errors that create variance
Lasts	Until the next receipt through the same defect	Until the process changes again
Right when	You need a verified opening balance or a financial count	Variances keep returning, or nobody can explain them

They aren't competitors. The strongest sequence is a data audit first, then fixes to the mechanisms it finds, then a count to reset the balance on a process that no longer rebuilds the error. Counting first and auditing never is the most common sequence, and the most expensive one over three years.

A worked example: the count that fixed everything for six weeks

A distribution center runs a wall-to-wall count in January. One item, a fast-moving consumable, counts 336 units short of the book across its reserve and pick locations. It's adjusted down, and the item closes the count with a clean record.

The item master says the case pack is 14. The vendor moved to 12 a year ago and nobody changed the field. So every case received posts two units that never came off the trailer. The receiver did nothing wrong, since the system does the conversion.

Through the year, the cycle count program kept finding the pick face short and kept correcting it, one adjustment at a time, each coded COUNT VAR. By January, most of the error had already been adjusted away piecemeal. The wall-to-wall count caught the remainder.

One item · 12 months · what the count saw vs what happened Illustrative arithmetic

Reconstruction of a case-pack error absorbed by count adjustments
Line	Units	Working
Cases received	1,200 CS	100 cases a month
Overstated by the pack field	2,400 EA	1,200 × (14 − 12), units posted that never arrived
Adjusted down by cycle counts during the year	2,064 EA	Many small adjustments, all coded COUNT VAR
Short at the January count	336 EA	2,400 − 2,064

What the count reported336 short · what the field generated2,400 units · and it restarts with the next receipt

The count was accurate and it was useless. It saw 336 units, a fraction of what the defect produced, and adjusted them away. Six weeks later the item is short again, because the pack field still says 14. A data audit finds this in two exports. Every receipt from that vendor leaves the same per-case gap, and the adjustment log shows the same item corrected downward again and again all year.

The same pattern is set out as a lookup in the root cause matrix (rows 03 and 11), and the pack mechanism in detail in unit-of-measure errors.

What an inventory accuracy audit should measure

If the audit's job is to tell you how accurate your inventory is, the first thing to audit is the accuracy measure itself. The same count file can honestly produce very different figures depending on three choices. The arithmetic is worked through in WMS inventory accuracy.

Net or absolute. Net accuracy lets overages cancel shortages, so a site with large errors in both directions can report a figure near 100%. Absolute accuracy sums the size of every miss. Only absolute accuracy predicts short picks.
Site, item or location level. Stock in the wrong slot is correct at site level and wrong at every pick. If pickers are shorting while the accuracy report is green, the report is measuring at the wrong level.
Units, value or line hits. Value-weighted accuracy is what finance needs. Line-level hit rate (the share of locations that counted exactly right) is what operations needs. Report both, and say which is which.
Tolerance. A tolerance band that treats a miss of a few units as a hit can make an accuracy figure look healthy while a recurring small error runs on every receipt. State the tolerance next to the number, and report the untoleranced figure alongside it.
Count timing. A count compared with the system quantity at post time rather than count time manufactures variances. Check the timestamps before trusting any accuracy trend. See cycle count discrepancies.

How to run a warehouse inventory audit

This sequence combines both kinds of audit. Steps 1 to 4 are the count, and 5 to 8 are the data work that makes the count worth doing.

Step 01

Define the question

A financial count needs a cut-off and full coverage. An operational audit needs coverage of the items and locations where the problems are. Write down which one you're doing. Mixing them produces a count that satisfies neither.

Step 02

Design the sample

For an operational audit, stratify: high-value and fast-moving items, then every item with repeated short picks or repeated same-direction adjustments in the last 90 days, then a random draw from the remainder as a control. The control group tells you whether the problems are concentrated or general.

Step 03

Count blind, and capture the system quantity at count time

Counters must not see the expected quantity. The system quantity has to be captured at the moment of counting, not when the count is posted. Exclude or recount any location with open tasks.

Step 04

Recount out-of-tolerance locations with a second counter

A variance that survives an independent recount is real. One that doesn't was a counting error, which is also worth knowing if it clusters by counter.

Step 05

Freeze the adjustments

Don't post the variances yet. The moment you adjust, the balance history stops being evidence. Hold them until step 7 has classified them.

Step 06

Replay the transaction history for the variances

For each confirmed variance, start from the last trusted balance and re-add every transaction in base UOM, doing the conversion yourself. The first row where your total and the system's part company is the event. The method is in inventory discrepancy.

Step 07

Classify and test for recurrence

Name the mechanism (duplicate receipt, pack error, unpaired move, keyed confirmation, interface drop, timing), then query the full history for the same signature. One instance is a correction. Forty is a finding.

Step 08

Adjust, with a reason code that names the cause

Now post the corrections. CASE PACK 12→14 VN-2140 is a record the next auditor can use. COUNT VAR is not.

What the audit report should contain

If you're buying an inventory audit, this is the list to hold the deliverable against. A report missing the last four items is a variance report, and your WMS can already print one of those.

Accuracy restated at location level, absolute, with tolerance stated, from the raw count file.
The variance list, confirmed by recount, with sign and size kept separate rather than netted.
Mechanisms, not categories. "Item master issue" is a category. "Case pack set to inner pack on 41 items from two vendors since March" is a mechanism.
Recurrence: how many times each mechanism appears in the history, not just in the count.
An owner for each finding, split between what your team can fix and what needs your WMS vendor or supplier.
An annual cost for each finding, from your own volumes and your own unit costs, used to rank the list.
The queries or logic, so the checks can be re-run next quarter without the auditor.

When to bring in outside help

Most of this can be done in-house by an inventory control lead with good export access and a few clear weeks. Few have the weeks. These are the situations where an outside data audit usually pays for itself:

Accuracy fell after a WMS go-live or upgrade and hasn't recovered. Migrated master data and changed defaults are the usual suspects. See WMS health check.
The same items keep appearing on the variance list after every count.
Write-offs are rising and they're being booked as shrink without anyone testing whether they're data.
Financial auditors or lenders are asking questions about the inventory figure that nobody can answer with evidence.
Pick shorts are high while reported accuracy is high. That combination nearly always means the accuracy measure is wrong.
Nobody internal is independent of the process. The team that designed the receiving flow is poorly placed to audit it.

The operational takeaway

Counting is how you find out how wrong the record is. It isn't how you make it stop being wrong. If your last physical inventory was accurate and the variances came back anyway, you've already proven the count isn't the fix.

Start with the WMS audit checklist to see which controls are missing. Then use the root cause matrix to name the mechanisms behind the variances you already have. If the list is longer than your team's time, that's what the audit is for.

Questions

How often should a warehouse inventory audit be done?: Most operations run a continuous cycle count program and a full physical inventory once a year, often because their auditors or lenders require it. A data audit of the records behind the counts is worth doing annually, and immediately after a WMS go-live, a large write-off, or a sustained fall in count accuracy.
Can we audit inventory without shutting the warehouse down?: A data audit needs no shutdown at all; it works from exports. A count audit can run live if counts are blind, the system quantity is captured at count time, and locations with open tasks are excluded or recounted. A wall-to-wall count for financial purposes usually does need a cut-off and a freeze.
What does a warehouse inventory audit cost?: Counting services are generally priced on labor, so cost follows SKU count, location count and hours on site. A data audit is priced on scope. WMSAudit's fixed fee is $1,500 to $2,500, quoted before work starts, and doesn't change if the data turns out to be harder than expected.
Is an inventory accuracy audit the same thing?: Usually it means the same job with the emphasis on the measurement: is the accuracy figure you report true, and at what level? A good one recomputes accuracy from the raw count file at location level, in absolute terms, and then explains the misses rather than just reporting them.