Cycle counting vs a full stock count: what each finds and how often to count
There are two ways to find out whether the stock on your shelves matches the stock in your system. A full stock count (a wall-to-wall or physical inventory) stops the building, counts every location once and gives you one clean snapshot. Cycle counting counts a small slice of locations every working day, so every item gets counted several times a year while orders keep shipping. This article compares what each one finds, works out how many counts a day a 2,000-SKU warehouse needs on three ABC schedules, prices both methods in labour hours, and ends with the cycle-count checklist we use.
In short: in WERC's 2019 DC Measures survey the median warehouse had 98.4% inventory count accuracy by location, and best-in-class was 99.9% or better; WERC's 2025 report puts the best-in-class line at 99.5%. In a 2,000-location warehouse, 98.4% means about 32 locations are wrong at any moment. An ABC cycle-count plan of A monthly, B quarterly, C yearly needs 33 counts a day (about 1.2 hours of one person); a full count of the same building takes about 160 hours in one push. The labour is close: about $6,700 a year against $5,600 a count. What differs is what you learn and when.
What each method finds
| Full stock count | Cycle counting | |
|---|---|---|
| How often each location is counted | Once a year (some sites twice) | Several times a year, more for fast or costly items |
| Operations during the count | Receiving and shipping stop, usually 1 to 3 days | Keep running; only the locations being counted are frozen for minutes |
| What it gives you | One valued snapshot of everything on one date | A running accuracy figure, week by week, by zone and by item class |
| Errors it finds | Every wrong balance on the day, all at once | Wrong balances within weeks of the mistake, while people still remember the receipt or pick |
| Root causes | Hard: an error found in December may date from March | Easier: recent errors trace back to a receipt, putaway, pick or unit-of-measure mix-up |
| Stock outside any location (in aisles, on docks, in returns) | Found, because every space is swept | Missed unless you add empty and odd locations to the plan |
| Who counts | Everyone, often with temps, on overtime or a weekend | One or two trained counters, in normal hours |
| Accuracy between counts | Drifts downward until the next full count | Measured continuously, so drift shows up and gets fixed |
The full count's one real strength is completeness on a single date: it sweeps every corner, including stock sitting where the system has no location for it. That is why sites keep one when records are poor, after a system change or a move, or when lenders or auditors insist on it. Cycle counting's strength is speed of discovery: an error caught three weeks after it happened can be traced and the process fixed, and an error caught eleven months later usually can't.
The accuracy benchmark: WERC's bands
WERC's DC Measures survey reports inventory count accuracy by location in five bands. It means the share of locations counted where the physical quantity matches the system quantity. Here are the 2019 bands, the last year the free handout printed the full row, with what each band means in a 2,000-location building:
| WERC band (2019) | Accuracy by location | Wrong locations out of 2,000 |
|---|---|---|
| Major opportunity | below 90.0% | more than 200 |
| Disadvantage | 90.0% to under 97% | 61 to 200 |
| Typical | 97% to under 99.012% | about 20 to 60 |
| Advantage | 99.012% to under 99.9% | about 3 to 19 |
| Best-in-class | 99.9% or better | 2 or fewer |
| Median | 98.4% | about 32 |
The 2025 DC Measures report, as summarised by Hyster-Yale, sets best-in-class for the same metric at 99.5% or better, which is 10 or fewer wrong locations out of 2,000. Either way, the gap between median and best is 20 to 30 wrong locations, and each one is a pick that fails, a substitute, a back order or an emergency buy.
Working out your own figure: take the locations counted in a period and divide the ones that matched by the total. Say the counters checked 33 locations a day for 20 working days, 660 counts, and 9 were off: 651 / 660 = 98.6%, inside WERC's typical band. Decide up front what counts as a match. A common pattern is zero tolerance on A items, and a small percentage or dollar tolerance on B and C items, so a 1-unit difference on a box of 1,000 screws doesn't count as an error while a missing laptop always does.
ABC cycle counting for a 2,000-SKU warehouse, worked
ABC counting splits items by annual dollar usage (units used a year times unit cost) and counts the top of the list most often. The usual split is about 20% of SKUs as A items (often around 80% of the dollar usage), 30% as B and 50% as C. For 2,000 SKUs that is 400 A, 600 B and 1,000 C items. To keep the arithmetic plain we assume one location per SKU and 250 working days a year.
Counts a day = (A SKUs × A counts a year + B SKUs × B counts + C SKUs × C counts) ÷ working days. Three schedules that warehouses commonly use:
| Schedule | A (400) | B (600) | C (1,000) | Counts a year | Counts a day |
|---|---|---|---|---|---|
| Light: A monthly, B quarterly, C yearly | 12 × 400 = 4,800 | 4 × 600 = 2,400 | 1 × 1,000 = 1,000 | 8,200 | 32.8 |
| Middle: A every 2 weeks, B every 2 months, C twice a year | 26 × 400 = 10,400 | 6 × 600 = 3,600 | 2 × 1,000 = 2,000 | 16,000 | 64 |
| Tight: A weekly, B monthly, C quarterly | 52 × 400 = 20,800 | 12 × 600 = 7,200 | 4 × 1,000 = 4,000 | 32,000 | 128 |
Note how the A items drive the load: on the light schedule they are 20% of the SKUs but 59% of the counts (4,800 of 8,200). On the tight schedule they are 65%. If you have to cut counts, the C frequency is the cheap lever and the A frequency is the one that protects most dollars.
A few counts are added on top of the plan, whatever the schedule: a location that hits zero or goes negative in the system, a pick that came up short, a location a picker flagged, and every location touched by a recount. Many sites also count an item when its system quantity is lowest (just before replenishment), because there is less to count.
What the two methods cost in labour
Our assumptions, so you can swap in your own: a counter does 30 location counts an hour (two minutes each with walking and scanning), 5% of counts need a recount, and labour costs $23.33 an hour: the BLS May 2025 national median for stockers and order fillers, $37,330 a year or $17.95 an hour, plus 30% for taxes and benefits. Overtime is 1.5 times that.
| Method (2,000 locations) | Counts | Hours | Labour cost |
|---|---|---|---|
| Cycle counting, light schedule | 8,200 a year + 5% recounts | 287 a year (1.15 a day) | $6,696 a year |
| Cycle counting, middle schedule | 16,000 a year + 5% recounts | 560 a year (2.24 a day) | $13,066 a year |
| Cycle counting, tight schedule | 32,000 a year + 5% recounts | 1,120 a year (4.48 a day) | $26,131 a year |
| Full count, two blind counts per location | 4,000 + 100 recounts | 133 + 3 + 24 for supervision and reconciliation = about 161 | $5,624 a count on overtime |
On labour alone, one full count a year and the light cycle-count schedule cost about the same. Three things the table leaves out tip the balance. First, the full count stops the building: a site that ships 400 orders a day either delays them or works extra days to catch up, and receiving waits too. Second, the full count brings in people who count rarely, which is where miscounts come from; that is why it is done twice. Third, the cycle count's hours mostly fit into slow parts of normal shifts, so the real extra cost is often lower than the table shows. Going from the light to the tight schedule quadruples the cost, so the tight one only pays where A items are costly or stock-outs are expensive.
Can cycle counting replace the year-end count?
Often, yes, if the auditors agree. The US auditing standard for inventories, PCAOB AS 2510, says in paragraph .10 that when well-kept perpetual records are checked by the company periodically against physical counts, the auditor's observation can be done during or after the end of the year, rather than at one wall-to-wall count on the last day. In practice the auditor will want to see the count plan, the accuracy records and how differences were investigated, and will observe some cycle counts. A program that reaches WERC's advantage band and can show its records is the usual starting point for that conversation. Keep a full count when records are below the typical band, after a system migration or a move, after a theft or flood, or when a lender requires it.
The cycle-count checklist
Before the count: today's locations come from the count plan (A, B, C due plus the triggered ones: zero, negative, short-pick, flagged). Receipts and picks into those locations are finished or held for the count. Counters get a blind count sheet or scanner task that does not show the system quantity.
During the count: count in the unit the system uses (eaches, not cases, unless the case is sealed and the system counts cases). Count damaged and expired stock apart. Check the item number on the label, not just the location, since a wrong item in the right slot is a classic error. Note stock found outside a location.
Differences: anything outside the tolerance for its class gets a second, blind count by a different person. Before adjusting, check open transactions: unposted receipts, picks in progress, transfers, returns. Only then post the adjustment, with an approver for anything above a dollar limit.
Root cause: for every confirmed error, record the cause (receiving count, putaway to the wrong slot, pick from the wrong slot, unit-of-measure mix-up, unrecorded damage or move, theft). Fix the process, not just the number. Track accuracy by zone and class each week against your target band.
Our cycle count checklist has those steps as items with sections. Press Start, set the zone, and send it to the counters; the person who reviews differences gets the photos and notes in one place. Picking errors are a big source of the differences a cycle count finds: our piece on what a mis-pick costs and which checks catch it works through that side, and the forklift daily inspection, item by item covers the other daily list most warehouses run.
Start the cycle count checklist: today's locations, blind counts, second counts on differences, adjustments and root causes, ready to change and send to your counters.
Open the cycle count checklistSources
WERC, DC Measures 2019 Snapshot in Time (handout), table row "Inventory Count Accuracy by Location": below 90.0%, 90.0 to under 97%, 97 to under 99.012%, 99.012 to under 99.9%, 99.9% or better, median 98.4. WERC 2019 handout PDF
Hyster-Yale, Top 12 distribution center metrics, from the 2025 WERC DC Measures Report: inventory count accuracy by location, best-in-class 99.5% or better. yale.com PDF
PCAOB, AS 2510 Auditing Inventories, paragraph .10 (observation where perpetual records are checked by periodic counts). pcaobus.org AS 2510
US Bureau of Labor Statistics, Occupational Employment and Wage Statistics, May 2025, stockers and order fillers (53-7065), national median $37,330 a year, as listed by WageDex. bls.gov OEWS
DCL Logistics, cycle count codes (A monthly, B quarterly, C annually), one published example of the light schedule. support.dclcorp.com
Worked counts, hours and costs: our own calculations from the assumptions stated in the article (2,000 SKUs, one location each, 20/30/50 ABC split, 250 working days, 30 counts an hour, 5% recounts, 30% labour on-cost).