By Luis Lopez, AI transportation consultant, CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast
Every warehouse says its inventory is accurate. What matters is how that accuracy is produced and how it is measured. In a third-party warehouse, where the goods belong to someone else and the system record drives orders, billing and reporting, the method behind the number is more important than the number.
Cycle counting is that method. It is the ongoing practice of counting a portion of the inventory on a schedule, comparing the count to the system record, and correcting both the record and the cause of any difference. Done well, it keeps stock records reliable without shutting the building down. Done poorly, it becomes a routine that adjusts numbers and learns nothing.
Cycle counting versus a full physical inventory
A full physical inventory counts everything at once. Receiving and shipping usually stop, the whole team counts, and the system is reset to what was found. It gives a clean snapshot, but it is disruptive, it happens rarely, and accuracy starts drifting again the next morning.
Cycle counting spreads the work across the year. A set of locations or items is counted each day or each week while operations continue. Errors are found closer to when they happened, which makes the cause easier to trace. Many operations use both: cycle counts through the year, with a full count when a customer, auditor or contract requires one.
How warehouses decide what to count
There are several accepted ways to build a count schedule, and most 3PLs combine them.
- ABC counting. Items are ranked by value, movement or both. The A items, which carry the most value or turn the fastest, are counted most often. C items are counted less often.
- Location-based counting. The counter works through the building aisle by aisle so every location is visited within a set period, whether or not it is supposed to hold stock. This catches product sitting where the system says nothing exists.
- Opportunity counts. A count is triggered by an event: a location that should now be empty, a pick that came up short, or a bin that falls below a set quantity. Low-quantity counts are fast and very accurate.
- Exception counts. Any discrepancy reported by a picker, a customer or a receiving clerk generates a count of the affected item and nearby locations.
The schedule should reflect risk. Small, high-value, easily confused or fast-moving items deserve more attention than full pallets of a single slow item in reserve racking.
What a count actually involves
The mechanics matter because small process choices decide whether the count is trustworthy.
- The task is issued. The warehouse management system assigns a location or item to a counter, usually on a handheld scanner.
- The count is blind. The counter is not shown the expected quantity. If the screen shows the system number, people tend to confirm it rather than count.
- The counter records what is there. Item, lot or serial number if tracked, quantity and unit of measure.
- The system compares. If the count matches or falls within a set tolerance, the task closes.
- Variances are recounted. A second person, ideally someone who did not do the first count, recounts before anything is changed.
- The cause is investigated. Recent receipts, picks, moves and adjustments for that item and location are reviewed.
- An adjustment is approved. Someone with authority signs off, and a reason code is recorded.
Timing is the detail most often missed. If a location is counted while a pick or putaway is in progress, the count and the system will disagree for reasons unrelated to any real error. Good operations lock the location or count only when open transactions are cleared.
The accuracy number and what it hides
When a warehouse quotes an inventory accuracy figure, ask how it is calculated. There are at least three common definitions, and they produce very different results from the same building.
- Location accuracy: the share of counted locations where item and quantity matched the record exactly.
- Unit accuracy: total units found compared with total units expected. Overages in one place can cancel shortages in another, so this figure tends to look better.
- Value accuracy: the same idea measured in inventory value. This is the most forgiving measure because gains and losses net out.
A warehouse can report excellent net unit accuracy while a meaningful share of its locations are wrong. For order fulfillment, location accuracy is the one that matters, because a picker sent to a bin needs the right item to be in that bin. As an AI transportation consultant, the first thing I look at in a warehouse scorecard is which of these definitions is being used and whether the absolute value of variances is reported, not just the net.
Where variances come from
Product rarely disappears. Most variances trace back to a process error that was recorded wrong at the time.
- Receiving errors: the wrong quantity or wrong item was keyed in at the dock
- Unit of measure confusion between eaches, inner packs and cases
- Putaway to a location other than the one scanned
- Picks from the neighboring bin, which creates a shortage in one item and an overage in another
- Similar-looking items or labels
- Returns placed back into stock without a system transaction
- Damaged goods set aside but never moved to a hold status
- Kitting or relabeling work that was not fully transacted
This is why root cause work is the real value of cycle counting. An adjustment fixes one record. Finding out that a particular receiving step keeps producing unit of measure errors fixes hundreds of future records.
What a shipper should ask its 3PL
If your goods sit in someone else’s building, you are entitled to understand how they are counted. Useful questions include:
- How often is each of my items or locations counted, and what triggers an extra count?
- Are counts blind, and who performs recounts?
- What tolerance closes a count without investigation?
- Who approves adjustments, and will I see every adjustment with a reason code?
- How is the reported accuracy figure calculated?
- How are shortages handled financially under the agreement?
That last point belongs in the contract, not in a conversation after a loss. Many warehouse agreements include a shrinkage allowance and a limit on the warehouse’s liability per unit or per pound. Understand those terms before inventory goes in. Two related reads are warehouse liability for lost and damaged goods and the questions to ask before signing a warehouse contract.
Technology that supports counting
Barcode scanning and a properly configured warehouse management system are the foundation. Beyond that, several categories of tools can extend counting capacity: RFID for item-level reads without line of sight, camera-based systems and drones that scan pallet labels in high racking, and analytics that rank locations by the likelihood of error so counters are sent where problems are most probable.
These tools are worth evaluating, with realistic expectations. A camera can confirm that a labeled pallet is in a location. It cannot confirm how many units are inside a closed case. Predictive models are only as good as the transaction history behind them. For a wider view of where automation fits, see the top uses of AI in warehousing and 3PL operations.
The takeaway
Inventory accuracy is not a one-time result. It is the output of a disciplined routine: count on a risk-based schedule, count blind, recount before adjusting, record why each variance happened and fix the process that caused it. A warehouse that can show you that routine, with the adjustment history to back it up, is giving you something far more useful than a single percentage.
For more practical explanations of how warehousing and freight operations really work, subscribe to the Freight Guru Podcast.
About the author: Luis Lopez is a Miami-based AI transportation consultant and logistics entrepreneur, the CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast.