Most small warehouses still run inventory the same way: once a year, shut everything down, count every box, and fix the system to match. The system is wrong again within weeks. If you've ever oversold a SKU you were sure was in stock, or found a pallet that "didn't exist," you've seen why. There is a better way, and it doesn't require stopping a single shipment.
Cycle counting counts a small slice of your inventory every day or week, so each SKU is counted several times a year while the warehouse keeps running. A yearly physical count gives one accurate snapshot and then decays. For most ecommerce operations, cycle counting is cheaper, faster, and more accurate, and it fixes the process errors that cause bad counts in the first place. Keep the full physical count for migrations, moves, and audits.
Why the Annual Physical Count Keeps Failing
A wall-to-wall count is a snapshot. It's accurate for roughly one day, and every receiving mistake, mis-pick, and unrecorded return after that starts pulling the system away from reality again. Because nobody looks until next year, errors compound for twelve months.
Shipping and receiving typically pause for a day or more, which means missed carrier pickups and delayed orders.
Counting everything in one push, often with temporary help, produces its own errors.
You learn that the numbers were wrong, but not when or why, so the same errors repeat.
Eleven months of drift between counts means you plan, buy, and promise stock on numbers you can't trust.
How Cycle Counting Works
Instead of counting everything once, you count a defined set of locations on a schedule. Each count compares the physical quantity to the system quantity, and any difference triggers a recount and an investigation. Over a quarter or a year, every location gets covered.
-
Choose what to count today
Pick the day's locations using one of the methods below, usually 20 to 50 locations for a small operation.
-
Count blind
The counter sees the location and SKU but not the system quantity, so the number isn't nudged toward the expected answer.
-
Recount any variance over your threshold
A second counter, or the same counter on a second pass, confirms the discrepancy before anything is adjusted.
-
Find the cause, then adjust
Trace the variance: wrong receiving quantity, mis-pick, put-away to the wrong bin, unit-of-measure error. Only then correct the system, with a note on why.
-
Track accuracy weekly
Report the share of counted locations that matched. A rising number means the root-cause work is paying off.
The Four Cycle Count Methods
You don't need all four. Most operations start with ABC and add a trigger-based count on top.
Count high-value, fast-moving SKUs most often and slow, cheap ones least. Matches effort to risk.
Walk a set of aisles each day so the whole building is covered on a fixed rotation. Easy to run on paper.
Count a location when something suggests a problem: a zero or negative balance, a short pick, a large receipt.
Count a random sample and extrapolate accuracy. Useful for measuring, less so for fixing individual SKUs.
A common ABC starting point: count A items monthly or quarterly, B items two to three times a year, and C items once or twice a year. Then let your own error history adjust the schedule, with problem SKUs counted more and clean ones less. If you haven't ranked your SKUs by velocity yet, the slotting section of How to Improve Warehouse Efficiency walks through the ranking.
Cycle Count vs Physical Inventory: Side by Side
The physical count's one advantage is that it's complete and auditors know it. Cycle counting beats it on nearly everything else: no downtime, smaller counting teams, faster detection of errors, and a trail that points to the process causing them. Keep the full count for a system migration, a facility move, onboarding a new 3PL, or whenever an auditor or lender requires it.
Where Bad Counts Actually Come From
Counting more often only helps if you fix what the counts reveal. In our experience these are the usual sources of variance, roughly in the order we find them:
- Receiving errors. The wrong quantity or SKU is booked in, so every downstream number is wrong from the start.
- Put-away to unrecorded locations. Stock is physically there but not where the system says.
- Unit-of-measure mix-ups. Eaches versus cases versus inner packs.
- Mis-picks. The wrong item or count leaves the building and one SKU goes short while another goes over.
- Unprocessed returns and damage. Items sit in a returns cage, or are thrown away, with no system adjustment. See the real cost of ecommerce returns for how much leaks out here.
- Unlogged manual adjustments. Someone fixes a number without recording why, destroying the audit trail.
A count program that adjusts the system without tracing causes just paints over the problem. The value is in the variance review, not the counting.
Setting Up a Program in Two Weeks
-
Week 1: clean the baseline
Make sure every location has a label, every SKU has a barcode or clear ID, and you have a way to export on-hand by location. If accuracy is badly off, do one full count first as the reset.
-
Week 1: rank and schedule
Rank SKUs by value and pick frequency, set A/B/C count frequencies, and spread the counts into a daily list of manageable size.
-
Week 2: set rules
Define the variance threshold that triggers a recount, who may approve an adjustment, and the reason codes required for each one.
-
Week 2: start, then review weekly
Begin counting and hold a short weekly review of accuracy and the top variance causes. Fix the process behind the biggest cause each week.
Before you expand your footprint because it feels like stock is missing, confirm it really is. Phantom shortages are often a counting problem rather than a space problem, as covered in 10 Warehouse Space Mistakes That Cost Sellers Money.
When to Hand Counting to a 3PL
If you don't have the staff or the discipline to run a count program alongside everything else, it's one of the easier tasks to outsource. A good 3PL counts as part of daily work and should be able to tell you its inventory accuracy rate, how often it counts, and what it does when it finds a variance. Those are worth asking in any evaluation, and they belong on the list in How to Choose a 3PL. At Simple Distribution we count inventory from our own warehouse in Selmer, Tennessee, and clients can ask for a count of any SKU when they need to trust a number before a promotion or a reorder.
Frequently Asked Questions
A physical inventory counts every item in the warehouse on one date, usually at year-end, and typically stops shipping and receiving while it happens. Cycle counting counts a small slice of inventory every day or week so that every SKU is counted several times a year, with operations running normally. A physical count gives you one accurate snapshot; cycle counting keeps accuracy high all year.
In most cases yes. Auditors generally accept a well-documented cycle count program as a substitute for an annual wall-to-wall count, provided the program covers all inventory, uses statistically sound sampling or full-coverage frequencies, and keeps a record of counts and adjustments. Confirm with your own accountant or auditor before dropping the annual count, because requirements vary by company and lender.
It depends on value and velocity. A common ABC pattern counts A items (high value or high velocity) monthly or quarterly, B items two to three times a year, and C items once or twice a year. Adjust the frequencies to your error history: SKUs that keep coming up wrong earn more counts, SKUs that are consistently clean earn fewer.
Best-run warehouses target 98 to 99 percent or higher at the SKU-location level. Many small and mid-size operations that have never run a formal program sit well below that. Measure accuracy as the share of counted locations where the system quantity matched the physical quantity, and track it weekly so you can see whether the program is working.
No, but it helps. You can run a program from a spreadsheet and a printed count sheet for a few hundred SKUs. Once you pass that, a WMS or inventory system that generates count tasks, records variances, and requires a recount above a threshold saves hours and removes manual errors. Barcode scanning matters more than the software brand.
The usual causes are receiving errors (wrong quantity or SKU booked in), put-away to an unrecorded location, picking the wrong item or quantity, unit-of-measure mix-ups such as each versus case, unprocessed returns, damage that is never written off, and manual adjustments with no audit trail. Cycle counting finds the symptoms; the root-cause review of each variance fixes the process behind them.
In a blind count the counter is not shown the system quantity, only the location and SKU. That prevents counters from nudging the number toward what the system says. Most programs use blind counts and trigger an automatic recount when the first count differs from the system by more than a set threshold.
The cost shows up as oversells and canceled orders, expedited replenishment, safety stock you carry to cover for distrust in the numbers, and write-offs when stock is finally found to be missing. The exact figure depends on your margins and volume, but most sellers who tighten accuracy are able to cut safety stock and cancellations measurably.
Yes. A good 3PL counts as part of normal operations and should give you accuracy reporting, variance explanations, and the ability to request a count of any SKU on demand. Ask any 3PL you are evaluating for their reported inventory accuracy rate, how often they count, and what happens when a variance is found.
After a system migration, after a move between facilities, when starting with a new warehouse or 3PL, when accuracy is so low that cycle counts cannot keep up, or when your auditor or lender requires one. Think of the full count as a reset button, and cycle counting as the maintenance that keeps you from needing it again.
Don't trust your inventory numbers?
Tell Matt how you count today. Simple Distribution will tell you honestly whether you need a better process in your own warehouse or a 3PL that already runs one.