A cycle count checks a small slice of inventory on a rolling schedule instead of counting everything at once. Done properly, it touches 20 to 60 SKU-location combinations a day, takes under an hour, requires no shutdown, and produces better accuracy than an annual wall-to-wall count because errors get found while the cause is still traceable.
The design decisions that make it work are all made before anyone counts anything: what gets counted how often, when the cutoff falls, what variance triggers what response, and who is allowed to post an adjustment.
Counting everything once a year finds the total error and destroys the information about where it came from. A 340-unit shortfall discovered in January could have originated in any of the previous twelve months, from receiving, from picking, from a mis-scanned transfer, or from theft. By the time it surfaces, the trail is cold and the only available response is an adjusting entry.
Shrinkage in retail generally is not trivial. The National Retail Federation's National Retail Security Survey 2023 found the average shrink rate rose to 1.6 percent of sales in fiscal 2022, up from 1.4 percent the prior year, which the NRF put at $112.1 billion across U.S. retail. NRF has since paused that annual report, so the figure should be read as the last full published benchmark rather than a current reading.
For a seller doing $8M in sales, 1.6 percent is $128,000. Finding that once a year as a lump is very different from finding it in fifty small pieces with a traceable cause attached to each.
Not everything deserves the same attention. Rank SKUs by annual cost of goods sold, then split them.
A items. The top 20 percent of SKUs by annual COGS, which typically carry 70 to 80 percent of the value. Count monthly.
B items. The next 30 percent. Count quarterly.
C items. The remaining 50 percent, carrying a small share of value. Count twice a year.
Then override the classification for anything with a history of variance, anything with high theft appeal, and anything with a short shelf life or a rapid obsolescence curve, regardless of value. A $4 accessory that goes missing constantly deserves A-item attention even though it will never rank on cost.
Take the count obligations and spread them evenly across working days.
For a 640-SKU catalog across three locations, the arithmetic works out like this:
Total: roughly 540 count events per month, or 27 per working day. At two minutes each including travel and system entry, that is under an hour a day for one person.
That is the number worth checking before committing to a program. If the daily count load exceeds ninety minutes, either the frequencies are too aggressive or the catalog needs to be pruned.
This is the step that gets skipped and the reason most counts produce variances that are not real.
Count before the day's picking starts or after it ends. Never mid-flow.
Freeze the SKU-location combinations being counted for the duration of the count. Most warehouse systems support a count freeze. If yours does not, count first thing.
Handle in-transit explicitly. Units that left the origin and have not been received belong to neither location's on-hand figure and should not appear as a shortage at either end. This is why an in-transit state matters so much for count integrity, a point covered further in the inventory in transit glossary.
Decide how open orders count. A picked-but-not-shipped unit is physically present and logically committed. Pick one convention, write it down, and apply it every time.
Do not show the counter the expected quantity. A counter who knows the system says 412 will find 412.
The counter records what they see. The system compares. Anything outside tolerance gets a second count by a different person before any adjustment is posted. That second count is where most apparent variances disappear.
Two thresholds, not one.
A unit threshold for small counts. A 3-unit variance on a bin of 12 is worth investigating. A 3-unit variance on a bin of 4,000 is noise.
A value threshold for the dollars. Any variance above a set dollar amount at landed cost gets a root cause investigation regardless of unit count.
A workable pair for a mid-size seller: recount anything outside 2 percent of expected quantity, and investigate anything above $250 at landed cost.
The investigation matters more than the adjustment. Categories to look for: receiving error on a partial pallet, mis-scan at pick, unrecorded transfer, damaged units removed without a write-off entry, and returns restocked without a system entry. Each has a different fix, and none of them are fixed by posting an adjustment.
An inventory adjustment is two entries, not one. Quantity moves and value moves.
Take a variance of 62 units short at a landed cost of $9.40. Inventory decreases by $582.80. The offset does not go to cost of goods sold as a lump, or it will contaminate your gross margin analysis. It goes to an inventory shrinkage or adjustment account that sits within or adjacent to COGS but is separately visible.
That separation is what lets you answer the question a lender or a buyer will ask: how much of your cost of goods sold last year was actual product sold, and how much was loss.
A seller runs the program above for a month. 538 count events completed.
Of the 23 recounts, 9 resolved to the system quantity, meaning the first count was wrong. The remaining 14 were real, totaling 388 units short and 96 units over, a net 292 units short at an average landed cost of $7.85, or $2,292.20.
Root causes on the 14: six receiving errors on partial pallets from one supplier, four mis-scans at a single pick station, two unrecorded transfers, one damaged pallet disposed of without an entry, one theft-consistent pattern on a small high-value SKU.
The $2,292.20 adjustment is the small part. The six receiving errors from one supplier are worth a conversation that prevents the next twelve. The four mis-scans at one station are worth an afternoon of retraining.
Stock tracked by location so a count is scoped to a place rather than a total. An in-transit state so units in flight do not read as shortages. Costing applied per transaction so an adjustment can be valued at the correct layer. And inventory reports by warehouse, value, and aging so the count results can be checked against a value that means something.
Those are the mechanics ConnectBooks provides for multichannel sellers, with FIFO valuation and stock tracked at the warehouse level. Note that it tracks by warehouse rather than by bin or zone, so bin-level count scoping stays a warehouse management system job.
They run the count, post the adjustment, and stop. The count is a detection mechanism. The value comes from the root cause work afterward, which is unglamorous and is the only thing that changes next month's result.
Track two numbers over time: the percentage of counts that match exactly, and the total value of adjustments as a share of COGS. If both are improving, the program is working. If adjustments keep landing at the same rate, you are measuring a problem rather than fixing it, and the answer lives upstream in receiving, picking, or how returns and settlements post to your books before it shows up in margin by SKU.
Running an e-commerce business comes with plenty of challenges, but ConnectBooks is here to make your life easier. With real-time insights, seamless integrations, and detailed tracking of your profitability and inventory, you can stay ahead of the game. Whether you’re selling on Amazon, Shopify, Walmart, TikTok or eBay, ConnectBooks helps you manage your finances with 100% accuracy and confidence, so you can focus on growing your business.
Ready to level up? Start making smarter, data-driven decisions every step of the way. Try ConnectBooks Free Today or Schedule a Demo
