Your system says you have 40 units. The shelf has 28. Somewhere, 12 units vanished, and nobody knows when. This is the quiet tax of poor inventory accuracy: stockouts that stall production, rush orders that eat margin, and month-end adjustments that make your numbers look like guesswork.
The traditional fix is the annual physical inventory: shut everything down for a weekend, count every SKU, find a pile of errors, then watch accuracy drift again within weeks. There is a better way. Cycle counting keeps your inventory records accurate all year, with no shutdowns. Here is how the methods work and which one fits your operation.
Why does inventory accuracy drift? #
Inventory records drift for ordinary reasons, not because anyone is careless:
- Receiving errors. Partial shipments, wrong quantities on packing slips, items put away in the wrong location.
- Picking errors. The right part in the wrong box, or two picks recorded as one.
- Untracked movements. Engineering borrows parts for a prototype, maintenance grabs a fitting, a return lands in the wrong bin.
- Transaction timing. The shipment went out Tuesday but the system posted it Thursday.
Each error is small. Hundreds of them, over a year, compound into records you cannot trust. When people stop trusting the system, they walk to the shelf instead of checking the screen, and the drift accelerates.
What is cycle counting? #
Cycle counting replaces the annual count-everything marathon with a planned schedule of small counts. Instead of counting 100% of SKUs once a year, you count a subset of SKUs every day or every week, so that every item gets counted several times a year and the warehouse keeps running the whole time.
The core metric is inventory accuracy, measured as the percentage of counted SKUs whose system quantity matches the physical count (within an acceptable tolerance). Best-in-class warehouses hold above 98%. If yours is below 90%, planning and purchasing are working on fiction.
Cycle counting works because it does two jobs at once: it corrects the quantity on record, and (more importantly) it forces you to investigate why the record was wrong. The count is the detection; the root-cause fix is the cure.
Which cycle counting method is right for your warehouse? #
There are four established methods. Most warehouses combine two of them.
1. ABC cycle counting (the most common) #
Classify SKUs by value or movement into A (high value or fast movers, roughly the top 20% of SKUs), B (middle 30%), and C (the remaining 50%). Then set count frequencies: A items monthly, B items quarterly, C items once or twice a year.
2. Control group method #
Pick a small, representative set of SKUs (say 50 to 100) and count them every day or every week. Use the group’s accuracy as a live proxy for the whole warehouse.
Why it works: it is the fastest way to surface systemic problems, such as a receiving process that consistently over-counts. It is also the best method for proving to skeptics that a problem exists before you invest in a full program.
3. Opportunity-based counting #
Count an item when it is already in your hands: at receiving, when it is picked for an order, when it is moved between locations, or when it hits a reorder point.
Why it works: it costs almost nothing because no one makes a special trip. A picker counting units during a pick adds seconds and catches errors at the exact moment they happen.
4. Random sample method #
Count a random selection of SKUs each day. Because the sample is statistically representative, it gives an unbiased estimate of overall accuracy.
Why it works: it prevents the bias of always counting the “easy” or “known-problem” items. Auditors like it for the same reason: it cannot be gamed.
How do you investigate and fix count variances? #
Counting without investigation is just expensive arithmetic. Every variance beyond tolerance needs a short, disciplined follow-up:
- Recount first. Half of all variances turn out to be counting errors: wrong unit of measure, a neighboring location counted by mistake, or a pallet tucked behind the racking. Recount before you touch the books.
- Check recent transactions. Was there a receiving or pick for this SKU in the last 48 hours? Was the transaction posted at all?
- Check the location. Is the item stored where the system thinks it is? Duplicate location entries (the same item listed in two bins) are a classic source of phantom variances.
- Fix the cause, not the person. If a picker counted wrong because the label was smudged, the fix is new labels, not a lecture.
- Adjust the record, document the cause. Post the adjustment with a reason code. Reason codes turn a pile of corrections into trend data: if “wrong location” keeps appearing, you have a put-away problem, not a counting problem.
What should your inventory accuracy target be? #
A realistic progression:
- Below 90%: crisis mode. Counts are mostly firefighting, and purchasing decisions are unreliable. Fix the big movers first with daily ABC counting.
- 90 to 95%: improving. Keep tightening root-cause fixes and expanding coverage.
- 95 to 98%: solid. This is where most well-run small manufacturers land, and it is good enough to plan production with confidence using your material planning data at https://artintech.ca/modules/material-resource-planning-software.
Measure accuracy as a rolling trend, not a one-time number. Post the weekly accuracy figure on the warehouse wall. When the whole team can see the number moving, it moves faster.
What to look for in cycle counting software #
Spreadsheets can run a count schedule, but they cannot fix the underlying causes. The right warehouse management system at https://artintech.ca/modules/warehouse-management-system turns counting from a chore into a control system. Look for:
- ABC classification built in. The system should auto-classify SKUs by value or velocity and generate count schedules from it.
- Mobile counting. Counts entered at the shelf via barcode scan beat paper sheets that get re-keyed (and re-typoed) at a desk.
- Variance workflow. Over-tolerance counts should trigger a recount and approval step automatically, with reason codes on every adjustment.
- Location-level accuracy. System quantities should live at the bin level, so variances point at a specific location, not a fuzzy warehouse-wide number.
That last point is the real payoff. Inventory accuracy is not a warehouse metric in isolation; it is the foundation everything else stands on. MRP plans, production schedules, and customer promise dates all assume the numbers are right. When they are, the whole operation gets calmer.
Start this week: classify your top 20% of SKUs, assign them a monthly count, and investigate the first five variances to root cause. You will find the first fix within days. And if your current setup makes even that simple schedule feel like extra work, it is worth a look at how integrated ERP for manufacturers at https://artintech.ca/solutions/cloud-erp-manufacturing-management-software handles counting, transactions, and planning in one place.
Want to see what barcode-driven cycle counting looks like inside a real ERP? Start a free trial of Artintech at https://artintech.ca/.
