GenFrontier homeScope a project
Menu
Reader questionWhen a cycle count doesn't match the system, how do we find the cause before posting an adjustment?

Investigating a cycle count variance before you adjust

How to investigate a cycle count variance in six steps (recount, unit of measure, cutoff, location, documents, ownership) before you post an adjustment.

field-notes / inventory-control / cycle-count-variance-investigation
A ball of string unrolled across wooden floorboards, its red thread running back to a nail head where it first snagged.

How to find the cause of a cycle count variance

Find the cause of a cycle count variance by testing the cheap explanations before the expensive ones, in a fixed order: a blind recount by a second person, a unit-of-measure check, a cutoff check for transactions still in flight, a search of other locations, a trace of recent receipts and shipments, and an ownership check. Stop at the first step that explains the full difference, fix the transaction that caused it, and post an adjustment only for what is left. The order matters because several of these causes correct themselves without any adjustment, and an adjustment posted against them creates a second variance later.

This field note walks through each step: what to look at in the ERP and the WMS, what a positive finding looks like, and what to do with it. It ends with the fields for a variance log and a rule for when to stop investigating. It does not cover which reason code to pick or who approves the adjustment. That is the subject of inventory adjustment reason codes and approval rules.

A short vocabulary first, so the rest reads the same way for everyone:

  • System quantity (also called the book quantity): what the ERP or WMS says is in the location at the moment of the count.
  • Variance: the counted quantity minus the system quantity. A positive variance means more on the shelf than the record shows. A negative variance means less.
  • Tolerance: the difference you accept as accurate for that item. The ASCM Supply Chain Dictionary defines inventory accuracy as the state “when the on-hand quantity is within an allowed tolerance of the recorded balance.”
  • Last good count: the most recent count of the same item and location that matched the record. The cause of today’s variance sits in the transactions between that count and this one.
The investigation order: cheapest test first, adjustment last 1 Blind recount, second person Count sheet, device log, scale calibration 2 Unit-of-measure test Pack conversions, receipt and pick UOM 3 Cutoff and in-flight transactions Open tasks, message queue, unposted docs 4 Search other locations Neighbors, staging, look-alike items 5 Trace receipts and shipments PO receipt, ASN, packing list, ship confirm 6 Confirm owner, lot and status Consigned, customer-owned, on hold No cause found: adjust as unexplained and count the item more often Cause found Fix the transaction: correct the receipt, post the move, or change owner, lot or status. Adjust only what is left.
Investigation order as described in this field note. Basis: ASCM Insights, "Inventory Discrepancy Inquiries" (2018), which recommends starting with the most likely or easiest-to-identify causes and recounting with a different person; the step sequence is GenFrontier's editorial synthesis.

Why corrections get posted and causes do not

Corrections get posted because a correction is one transaction and an investigation is several, and the count team is usually measured on finishing the count. The result is a clean record today and the same variance at the next count.

ASCM said this plainly in a 2018 article by Britt Lowry, CPIM: “If a discrepancy is discovered when users pick an item or perform a count, an inventory correction typically is made. Less often, actual root-cause analysis is performed to figure out why the discrepancy happened in the first place.” The article adds that the analysis is sometimes avoided because of “uncertainty about how to begin” (ASCM, Inventory Discrepancy Inquiries). A fixed investigation order removes that uncertainty.

The definition of cycle counting already assumes the investigation happens. The ASCM Supply Chain Dictionary (18th edition, 2024) says: “The key purpose of cycle counting is to identify items in error, thus triggering research, identification, and elimination of the cause of the errors.” A program that counts and corrects, but never researches, does only the first third of that sentence.

What the research says about record errors

Record errors are common, and they cluster. In a study of nearly 370,000 inventory records from 37 stores of one retailer, DeHoratius and Raman (2008) found 65 percent of the records inaccurate. They also found that inventory auditing practices were among the factors that reduce record inaccuracy, while a more complex store environment and distribution structure increased it.

That study looked at retail stores, not distribution centers, so treat the percentage as context rather than a benchmark. The mechanism carries over: stock moves, and the record of the move is late, wrong or missing.

A later paper by Best, Glock, Grosse, Rekik and Syntetos (2022) studied the less discussed direction: positive discrepancies, where more stock is on hand than the system shows. Its authors note that practitioners “generally focus more on negative IRI” (inventory record inaccuracy) because shortages cause availability problems. Overages matter too. An overage on one item is often the other half of a shortage on another.

Why benchmarks will not tell you if you are investigating enough

There is no single accuracy standard to compare against. ASCM measures inventory accuracy as “the percent of items with inventory levels that fall within tolerance” and says target values “usually are 95 percent to 99 percent, depending on the value of the item.” APQC’s open benchmark for inventory accuracy (measure 100781) is computed differently, from the variance between physical and perpetual inventory. In an archived copy of that page from January 2026, the cross-industry median was 95.0 percent on a sample of 8,660.

Glock, Syntetos and Rekik (2025) interviewed 25 retail executives and found “a wide variety of IRI metrics” in use and “little consensus on what constitutes an acceptable measure.” So define your own measure in writing, and judge the investigation program by whether the same causes stop coming back.

Freeze the evidence at the moment of the count

Before anyone investigates, capture what the system believed at the moment of the count and what has moved since. Without that snapshot, every later step compares the shelf to a moving number.

Many warehouses keep shipping and receiving while they count. Some systems do not lock a location during a cycle count. Microsoft’s documentation for Dynamics 365 Supply Chain Management states that when cycle counting work is created, “the on-hand inventory that is associated with this location isn’t blocked, and is therefore available for reservation and outbound processing, even though open counting work exists” (Microsoft Learn, Cycle counting). Other products behave differently. Find out what yours does, because it decides how much of step 3 you will need.

Capture these facts for every variance above tolerance, ideally automatically when the count is entered:

  1. Item, location, lot or serial number, license plate, owner and inventory status (available, hold, damaged, inspection).
  2. The system quantity at the moment of each scan, and which system it came from (ERP, WMS or both).
  3. The count timestamp, the counter’s ID, the device, the unit of measure the count was entered in, and whether the count was blind.
  4. Open tasks touching the location at count time: picks, put-aways, replenishments and moves, with their status.
  5. The date of the last good count for that item and location.
  6. The transaction history for the item and location since the last good count, from both the WMS and the ERP.

Item 5 bounds the search. If the location matched 12 days ago, the cause is somewhere in the transactions of the last 12 days. ASCM’s article makes the same point about timing: “The time to identify the root cause of an inventory discrepancy is right after it happens. There will be far fewer transactions to look through, and the people you talk to are more likely to remember if some related event happened.”

Read the shape of the variance before you walk the floor

The number itself often points to the step that will explain it. Before you send anyone to recount, look at the variance against the pack sizes, the open tasks and the neighboring locations.

Variance patterns and the step that usually explains them
What the variance looks likeSuspect firstStep
Recount by a second person disagrees with the first countCounting error, mixed or hidden cartons1
Exact multiple of a pack size (case, inner, pallet)Whole packs moved without a transaction, or counted in the wrong unit2
Exact multiple of a pack size minus one (for example 23 for a case of 24)Packs keyed as eaches, or eaches keyed as packs, on a receipt or pick2
Equals the quantity of a pick, put-away or move near the count timeCutoff: a transaction posted before or after the scan3
Equals a quantity waiting in the integration queue or error logWMS and ERP out of step3
Shortage here, equal overage in a nearby location or a look-alike itemWrong put-away, wrong pick or wrong replenishment4
Equals a receipt or shipment line quantity since the last good countReceipt or shipment posted wrong, or not posted5
Zero at item level but not at lot, owner or status levelLot swap, wrong owner, or stock in the wrong status6
Basis: the "multiple of the unit of measure" and "complementary discrepancies" tests come from ASCM Insights, "Inventory Discrepancy Inquiries" (2018). The other rows are working rules of thumb for the order of investigation, not measured frequencies.

Step 1: recount blind, by a second person

A recount by someone who did not do the first count, without seeing the expected quantity, rules out the cheapest cause: the count itself. ASCM’s advice is to “perform a recount of the inventory in question, preferably by someone who was not involved in the initial count.”

Blind means the counter does not see the system quantity. Microsoft’s documentation describes this as a design choice in the Dynamics 365 warehouse app: “The system never shows the expected quantity to count. This design prevents intentional miscounts.” The same app has a Number of attempts setting that lets a worker repeat a count, which is a useful first recount for small slips.

The second person matters as much as the blindness. Someone who counted a location once tends to miss the same hidden carton twice.

How to run the recount

  1. Count the whole location, not just the item that varied. If another item is in the location, it may be the missing stock under a wrong label.
  2. Open every partial case. Check behind, under and on top of pallets, and the floor below the location.
  3. Count in the base unit of measure, then convert. Do not let the recounter enter “4 cases” if a case holds a different quantity than the item master says.
  4. Check every license plate in the location against the system’s list of plates for that location.
  5. Record the recount as a separate entry with its own counter ID and timestamp. Do not overwrite the first count.

If you count with scales, the scale is part of step 1. ASCM: “If you use counting scales or other automated counting tools, ensure that they have been calibrated and that users know how to work with them properly.” A piece weight captured on the wrong sample size produces the same variance on every count of that item.

What the recount result tells you

Recount result What it means What to do
Recount matches the system The first count was wrong No adjustment. Log the counting error and its cause (hidden stock, label, scale, training)
Recount matches the first count The variance is real Continue to step 2
Recount differs from both The location is disorganized or mixed Third count with a supervisor, then go straight to step 4
Recount finds a different item in the location Misplaced stock Go to step 4 with both items

Many systems can enforce this. In Dynamics 365, a count outside the deviation limits on the Work users page (“Maximum percentage limit” or “Maximum quantity limit”) goes to a status of Pending review, and Microsoft notes that “the counted value is rejected if you don’t approve the difference.” Whatever your system, make sure a variance above tolerance cannot post from the handheld without a recount and a review.

Step 2: test for a unit-of-measure error

If the variance is an exact multiple of a pack size, suspect a unit-of-measure error before anything else. ASCM: “If the discrepancy is a multiple of the unit of measure, you most likely have a shipment or receipt that was made in the wrong unit type.”

Pull the item’s unit conversions from the item master: each, inner pack, case, layer and pallet. Then divide the variance by each pack quantity. A clean whole number is a strong lead.

The pack-minus-one test

There is a second pattern that the plain multiple test misses. When someone keys a quantity of packs into a field that expects eaches, or the reverse, the error per line is not the pack size. It is the pack size minus one.

For example, suppose a case holds 24 units and a receiver keys “9” into a field that defaults to eaches, when 9 cases arrived. The shelf gained 216 units and the record gained 9. The variance is 207, which is 9 times 23, not a multiple of 24. So divide the variance by the pack size and by the pack size minus one. Either clean result points to a unit-of-measure error.

Example (fictional). Item EX-30442, nitrile gloves, is stocked in boxes and packed 24 boxes to a case. The system shows 225 boxes in location B-14-02-C. Two counters both count 432. The variance is +207. It is not a multiple of 24, but 207 divided by 23 is exactly 9. The receipt history shows a purchase order line for 9 cases received six days ago, posted as 9 boxes. The fix is to correct the receipt to 216 boxes. That also corrects the purchase order’s received quantity, so the invoice match is right. No inventory adjustment is posted.

Where unit-of-measure errors enter

  • Receiving: the purchase order is in cases, the receiving screen defaults to eaches, or the supplier’s ship notice uses a different unit than your item master.
  • Item master changes: a supplier changes the case pack from 24 to 20 while old cases are still in stock, and the conversion is updated for everything at once.
  • Picking: a pick confirmed as “1” for a case pick that should have been 12 eaches, or the reverse.
  • Counting: the count is entered in cases while the system expects eaches. Some apps reduce this risk. Microsoft notes that when an item has several units enabled for cycle counting in Dynamics 365, “workers are asked to provide a quantity for each unit and the system adds these quantities together.”
  • Measured items: ASCM flags items “measured by length, weight or volume” as more prone to receipt unit errors. The dictionary’s own example of a purchasing unit of measure: “purchasing buys steel by the ton, but it may be issued and used in square inches.”

The retail research found the same mechanism. Practitioners in the Best et al. workshop said that “multiple single items entering the store could get registered as case packs, creating positive IRI.”

Fix the transaction, not the balance

When a unit-of-measure error sits on a receipt, an adjustment repairs the shelf balance but leaves the receipt wrong. The purchase order then shows the wrong received quantity, and the three-way match between order, receipt and invoice can approve the wrong amount. Correct or reverse the receipt and re-post it in the right unit. Then fix the item master or screen default that allowed the error. If the conversion table itself is wrong, every test in this step gives the wrong answer.

Step 3: check cutoff and transactions still in flight

A variance that equals the quantity of a pick, receipt, move or message posted just before or after the count is a timing difference, not a loss. The fix is to let the transaction finish, or to re-run it, never to adjust.

Cutoff is the line between transactions the system quantity includes and transactions it does not. A count compares a physical quantity at the moment of the scan with a system quantity at some moment. If those two moments differ, and stock moved in between, you see a variance that is not real.

ASCM’s checklist starts here, right after the recount: “Look for unposted transactions for the item in your enterprise resources planning (ERP) system. Ideally, shipping and receiving should not occur while you are counting; if they must, then be sure to account for them in your procedures.”

What to look for

  • Picks confirmed on the floor but not yet posted, or posted while the goods still sit in the location waiting for a consolidation move.
  • Put-aways scanned into the location after the snapshot, or put-aways posted to the location while the pallet still sits in the aisle.
  • Replenishment moves between reserve and the pick face during the count.
  • Messages between systems. If the count compares against the ERP, every WMS transaction that has not reached the ERP yet looks like a variance. Check the integration queue and the error log for the item.
  • Receipts and shipments waiting for paperwork. ASCM lists “parts that were put in inventory, but the receipt has not been posted. Perhaps the packing list was lost and not processed,” and “parts waiting to ship, but the sales transaction already has been posted.”
  • Work in process and backflush. The dictionary defines backflush as reducing component inventory automatically “based on what should have been used,” and warns that this approach “has the disadvantage of a built-in differential between the book record and what is physically in stock.” ASCM’s article adds: “If you use backflushing, material may have been physically put into WIP but not yet backflushed.”
  • Returns that were received but not posted, or posted but not received.

The echo: why an adjustment here makes things worse

An adjustment posted against an in-flight transaction does not just fail to fix the record. It creates a second variance of the same size, in the opposite direction, the moment the delayed transaction posts. If that second variance pushes the balance below zero, you also get the problem described in negative inventory in your ERP.

Example (fictional). Item EX-20118 in pick face A-03-02-B shows 240 units in the ERP. A counter scans the location at 9:20 and counts 216. The WMS task history shows a pick of 24 units from that location, confirmed on a handheld at 9:14. The WMS posted it at once, but the shipment confirmation to the ERP waits for the end of the wave and is still in the queue. If the inventory controller adjusts the ERP down by 24, the ERP will drop by another 24 when the wave closes. It would then show 192 against 216 on the shelf. The next count finds a variance of +24, and someone may post a second adjustment to cancel the first. The right action is no action: wait for the message, then confirm both systems show 216.

Make cutoff a rule, not a judgment

Write the cutoff rule into the count procedure. For example, suppose your rule is that any transaction touching the item and location within 30 minutes either side of the scan is reviewed before the variance is accepted. The window is yours to set. What matters is that it is written down and the system lists those transactions on the variance screen.

If the WMS directs the work but counts are checked against the ERP, the snapshot must come from the ERP, and the queue between them becomes part of every investigation. The design of that interface, and what happens when messages fail, is part of ERP and WMS integration work.

Step 4: search for the stock in the wrong location

If the recount stands and the units and timing check out, look for the stock somewhere else before you write it off. A shortage in one location is often an overage in another. ASCM’s own fallback is: “If all else fails, look for inventory of the item in other locations.”

The ASCM Supply Chain Dictionary has a term for the systematic version of this search. A location audit is “a methodical verification of the location records for an item or group of items in inventory to ensure that when the record shows an item’s location, it is, in fact, in that location.”

A search plan in order of likelihood

  1. Every other location the system says holds the item. Count them all. An overage there that matches the shortage here is a wrong put-away or a wrong pick location.
  2. Neighboring locations. Look at the levels above and below in the same bay, the bays on either side, and the matching location across the aisle. Also try transposed numbers, for example A-03-02 against A-02-03. A numbering scheme that people misread produces these errors every week, which is one reason to read how to number warehouse locations.
  3. Places stock waits. Check receiving staging, outbound staging, the quality hold area, the returns cage, overflow locations and trailers at the door.
  4. Look-alike items. Search for an equal and opposite variance on items with similar descriptions, adjacent item numbers, the same supplier, or a different size or color of the same product.
  5. Recent moves out of the location. Look for replenishments to a pick face, consolidation moves and put-aways where the operator overrode the suggested location.

Point 4 comes straight from ASCM: “You also should look to see if there are complementary discrepancies in a similar item. This might indicate that there is some confusion or a labeling issue for two parts with similar characteristics.” The same article notes that shipping a customer the wrong part “causes two inventory errors,” one on each item.

The retail simulation in Best et al. (2022) points the same way. The misplacement rate of items was “one of the main drivers of positive IRI,” and of the types of misplacement modeled, misplacement during shelf replenishment had “the largest effect.” In a distribution center the closest equivalent is the replenishment move from reserve storage to the pick face. That is the move to check first when a pick face is short and reserve looks heavy.

Example (fictional). A count finds item EX-41007 (work shirt, navy, size M) short by 6 in location C-11-03-A. A search of the system for recent variances on the same style finds EX-41008 (navy, size L) over by 6 in C-11-04-A, the next bay. Both cartons carry the same supplier label layout, with the size printed in small type. The fix is a move transaction of 6 units, plus a request to the supplier or the labeling station for a larger size field. No gain or loss is posted on either item.

Fix with a move, not two adjustments

When you find the stock, post a move from where the system thinks it is to where it is. Two adjustments (a loss here, a gain there) would net to zero at item level, but they would add two entries to your adjustment history and hide the real cause, a placement error. Log that cause on both items.

Step 5: trace receipts and shipments

When the stock is not in the building, the documents for what came in and what went out usually explain it. Compare each receipt and shipment since the last good count with the document the other party holds.

ASCM frames the search by the direction of the variance. When the system shows less than you counted, “you are looking for a transaction in your ERP system that doesn’t belong. This is one of the easier problems to find. The more difficult issue is identifying a transaction that was never created. This is when you have more inventory in your ERP system than you physically have on-hand.”

Causes by direction of the variance

The table collects the receiving and shipping questions from ASCM’s article, sorted by direction.

Direction Receiving side Shipping and returns side
Counted more than the system (positive variance) Stock put away but the receipt never posted, for example a lost packing list. Wrong item or wrong purchase order received for the same item. Receipt quantity or unit lower than what arrived Right item shipped against a sales order for the wrong item (two errors). Sales transaction posted while the goods still wait to ship. Short shipment posted as complete. Return received but not posted
Counted less than the system (negative variance) Receipt posted but goods not yet put away. Receipt posted from the packing slip for more than physically arrived. Wrong item received Wrong item shipped against the right pick list (two errors). More shipped than ordered. Shipment made but not yet posted. Return posted but not received. Samples pulled without a transaction. Expired stock scrapped without a transaction

Which documents to compare

  • Inbound: the purchase order line, the receipt transaction, the supplier’s packing list, the bill of lading, and the advance ship notice (X12 856, which ASC X12 names “Ship Notice/Manifest”). If the receipt was posted by scanning pallet labels against the ship notice, check whether the pallets were actually opened and counted.
  • Outbound: the pick, pack and ship confirmations and the carrier pickup record. A customer claim for a short or wrong shipment may be the other half of your count variance.
  • Third-party warehouses: compare your record with the warehouse’s messages. X12 names them 944 “Warehouse Stock Transfer Receipt Advice,” 945 “Warehouse Shipping Advice” and 947 “Warehouse Inventory Adjustment Advice.”

One receiving habit deserves its own mention. ASCM warns that an overstated receipt “can happen when someone takes receipt of an item based on the packing slip rather than counting the receipt quantity.” If your receiving process allows a receipt to be posted from the paperwork, expect negative variances at the next count, and expect them to look like theft.

Fix the document before the balance

A wrong receipt or shipment quantity is also a commercial error: you paid for, or billed, the wrong amount. Correct the document, raise the supplier claim or customer credit, and adjust only if the document proves right and the stock is still missing.

Step 6: confirm who owns the stock

Before you adjust, check that the stock you counted, or failed to find, belongs to you and sits in the status you think it does. Consigned stock, customer-owned stock in a third-party warehouse and stock on hold can make both the shelf and the record right, but in different buckets.

Ownership

The ASCM Supply Chain Dictionary defines consigned stocks as “inventories, generally of finished goods, that are in the possession of customers, dealers, agents, brokers, etc., but remain the property of the manufacturer by agreement with those in possession.” If you hold supplier-owned consignment stock, a count of the item can include units the system holds under a separate owner or a separate item. A count that ignores the owner field will show a variance that is really two balances added together.

The same problem appears in third-party warehouses that store the same product for more than one client. A receipt posted to the wrong owner makes one client’s stock look over and another’s short, with zero difference at the item level. Adjusting both would report a loss to a client who lost nothing. If the warehouse sends 947 adjustment messages to clients, that loss lands in the client’s ERP too. Our third-party logistics page covers owner-level inventory in more detail.

Lot and status

  • Lot or serial swap: the item quantity is right but the lot split is not. This happens when a picker takes the nearest lot rather than the allocated one. Correct the lot records with a lot reclassification, not a pair of adjustments, so traceability holds.
  • Status: a count compared with “available” stock will miss units in hold, damaged or inspection status. Check whether the variance equals a quantity in another status for the same item and location.

The fix in each case is an owner, lot or status transfer. No gain or loss is posted, because nothing was gained or lost.

The variance log and the decision to adjust

Record every variance above tolerance in a log, with the evidence, the step that explained it and the fix, even when no adjustment posts. After a few months the log shows which step finds most causes in your building, and which process keeps producing them.

ASCM’s article ends with the reason the log exists: “whatever the cause of an inventory discrepancy, make sure to close the loop by notifying the appropriate person so the problem is corrected. Once the responsible party has been made aware of the issue, keep monitoring the inventory to ensure that the corrective action is effective.”

Fields for a variance log

Field What to record Why it matters
Variance ID One ID per item, location and count Links the count, the investigation and any adjustment
Count time and counter Timestamp, counter ID, device, blind or not Cutoff checks and counter patterns
Item and location detail Item, location, lot or serial, license plate, owner, status Steps 4 and 6 depend on it
System quantity and source Quantity at the scan and its source system Shows what the count was compared with
First count and recount Both quantities and both counters Proves step 1 was done
Variance In base units and in value at standard or average cost Drives the depth of investigation
Last good count Date of the last matching count Bounds the transaction search
Steps completed Which of steps 1 to 6 were done and the result of each Shows where investigations stop
Cause found One cause category, such as count error, unit of measure, cutoff, misplacement, receiving, shipping, ownership, unexplained Feeds the trend review
Evidence Receipt, shipment, task and message IDs Lets an auditor or a colleague retrace the finding
Fix type Transaction corrected, move, owner or status transfer, adjustment Shows how many variances needed no adjustment
Adjustment and reason code Adjustment number and code, if one was posted Ties the log to the general ledger
Process owner and action Who was told, what will change, by when Closes the loop
Follow-up count Date and result of the next count of the item Proves the fix worked
Time spent Investigator minutes Tunes the rule for when to stop

When to stop investigating and post

Not every variance deserves all six steps. Set the depth of the investigation by value and by repetition, write it down, and apply it the same way every time. The thresholds below are hypothetical. Set yours with your controller.

Variance Investigation depth If no cause is found
Within tolerance None. Record the count as accurate Nothing to post, or post per policy
Above tolerance, under a hypothetical $250 Steps 1 to 3 Adjust with an “unexplained” code and count again sooner
Above a hypothetical $250, or the same item and location twice in 90 days All six steps Adjust with an “unexplained” code, approval at the next level, and raise the count frequency
Above a hypothetical $5,000 All six steps, with a second investigator Escalate before posting

When nothing explains the variance, the item needs more attention, not less. ASCM’s article describes the usual practice: “change the count frequency of an item with discrepancies to daily (or weekly for low-transaction-volume items) until the root cause has been identified and eliminated.” A shorter interval means fewer transactions to search next time. The reason code and approval rules for the adjustment itself are covered in inventory adjustment reason codes and approval rules.

What to measure monthly

Two numbers tell you whether the investigation program works: the share of variances closed without an adjustment, and the repeat rate (how often the same item and location, or the same cause, comes back). Track overall accuracy beside them, using one written definition. You can calculate accuracy by item and by value from a count file with the inventory accuracy calculator.

What to do before your next count

Write the six steps into your count procedure, make the system capture the evidence automatically, and start the variance log this week. The first month of the log will show which step closes most of your variances.

A checklist to work through:

  1. Decide which system the count is compared against, and take the snapshot from that system at each scan.
  2. Turn on blind counting and require a second-person recount before any variance above tolerance can post.
  3. Show pack conversions on the variance screen for the multiple and pack-minus-one tests.
  4. List open tasks and queued messages for the item and location within your written cutoff window.
  5. Add a “where else is this item” view with every location holding it and open variances on look-alike items.
  6. Link receipts, shipments and partner messages since the last good count, so step 5 is a query.
  7. Show owner, lot and status on every count line.
  8. Start the variance log and review it weekly by cause and location.
  9. Size the count workload with recounts included. The cycle count planner shows how many counts per day your class frequencies require.

Most of these items are screen and report changes, not new processes. If your ERP or WMS cannot show this evidence in one place, that gap is the kind of work described on our inventory management software page. In our engagements we start by writing down the count and investigation procedure before building anything, so the screens follow the procedure rather than the other way round.

Working on this problem?

If this field note describes a process you need to fix or build software around, tell us about it. We start with your process, your systems and your deadline.