Skip to main content
All articles

Blog

Count Day: Why Most of Your Year-End Inventory Difference in SAP Is Not Real

A year-end stock count shows a $31,240 shortfall, but only $7,410 is real. See where count day differences come from and how to check them against SAP before you post anything to the books.

Close-up of a count sheet on a clipboard with a corrected handwritten number in a warehouse aisle, while a second counter scans a bin label behind.

At 6:00 on the morning of Saturday, January 2, 2027, fourteen two-person teams spread out across the warehouse of a manufacturer of hydraulic hoses and fittings in Wisconsin. Each team carries a clipboard of count sheets printed from SAP the evening before, a pen and a handheld scanner that works most of the time. Two people from the external audit firm walk the aisles with them, picking bins at random and counting along. Production is stopped, the dock is closed, and nothing is supposed to move until the count is finished.

By 2:00 in the afternoon the teams have counted 3,412 items. The sheets come back to the inventory office in a cardboard box, and two clerks spend the rest of the weekend keying counts into SAP. On Monday morning the difference list is ready. 286 items do not match the book quantity in SAP. Valued at standard cost, the count is $31,240 short.

The plant controller has a choice that every controller faces at this point. She can post the $31,240 as an inventory loss, close the year and move on. Or she can find out what the number actually means before it goes into the books, with the auditors waiting and the close calendar ticking.

She chooses to look, and the team spends most of the following week finding out. A pallet of fittings was counted twice. A bin was empty because its stock had been put away one bin over. Brass fittings were counted in boxes while SAP holds them in single pieces. A count sheet had "48" crossed out and "84" written next to it, and the clerk keyed the first number. And 400 couplings used on a production order on December 31 never left the book, because the automatic goods issue failed and sat in an error list nobody had cleared.

When every explanation is in, the real loss is $7,410. More than three quarters of the $31,240 was paperwork.

The company and figures in this story are a composite, built to make a common pattern concrete. The pattern is real for almost every manufacturer and distributor that counts its stock at year end. Most of the difference on count day is not missing stock. It is a mismatch between what people wrote down, what SAP thinks happened, and when. The hard part of a physical inventory count is not counting. It is deciding which differences are real before you post them.

How a year-end count works in SAP

The SAP side of a physical inventory is well established, and most teams run it the same way. A physical inventory document is created for each storage location, often in bulk with MI31 or one at a time with MI01. It lists every material and batch to be counted, and the book quantity SAP will compare against. A posting block on the document stops goods movements for those items while the count is under way, and the book quantity can be frozen at the moment the document is created.

The documents are printed as count sheets with MI21 and handed to the count teams. The counts come back and are entered against each document with MI04. The difference list in MI20 shows every item where the count and the book disagree, with the value of each difference. Items that look wrong are recounted with MI11. When everyone is satisfied, MI07 posts the differences. Gains post with movement type 701 and losses with 702, and the value moves to the inventory difference account in the general ledger.

SAP also lets you set value limits per user for posting differences, so a clerk cannot write off a large shortfall without someone senior approving it. Most companies set a threshold above which the controller or the finance team has to sign off.

On paper, the process is tidy. In practice, three things make it messy.

The first is that people count on paper, or on scanners that export to spreadsheets, and those counts have to be typed into SAP by hand. Every typed number is a chance to get it wrong.

The second is that SAP compares the count with a book quantity frozen at a moment in time. Any movement that happened physically but was not posted before that moment, or was posted afterwards, shows up as a difference. Goods issues from production that failed, receipts physically put away but not yet posted, and returns sitting at the dock all look like missing or extra stock.

The third is that the difference list does not say why. It says that material HS-0816 in bin 02-11-07 is 36 reels short. It does not say that the count sheet says 84 and the clerk typed 48. Somebody has to go and find out, for each of the 286 lines, in the week everyone else is trying to close the year.

Where the $31,240 really came from

Go back to the Wisconsin plant and take the 286 differences one category at a time. This is the work the plant controller's team did by hand over five days.

Movements that had not been posted. 118 of the differences were timing, with a net value of $19,860 short. The largest was the 400 couplings at $14.20 each, used on production order 1004518 on December 31. The order was confirmed, but the automatic goods issue for its components failed and sat in the COGI error list. Physically the couplings were gone. In SAP they were still on the shelf. The count found $5,680 of "missing" stock that was in fact sitting inside finished hose assemblies. Other items had the opposite problem. Material received on the afternoon of December 31 was put away but not posted until January 4, so the count found more than SAP expected.

Units of measure and handwriting. 63 differences, $4,940 short, were reading errors. Brass fitting HF-2210 is held in SAP in single pieces, 1,200 of them. The count team counted 48 boxes of 25 and wrote 48. Typed into SAP, that became a shortage of 1,152 pieces, worth $2,131.20. On another sheet, reel hose HS-0816 had "48" crossed out and "84" written next to it in a different pen. The clerk keyed 48, and a shortage of 36 reels worth $2,232 appeared that did not exist.

Wrong bins and double counts. 68 differences netted out at $970 over. A full pallet of 240 fittings of HF-3340 at $26.75 each sat on the boundary between two count zones, and both teams counted it, creating a gain of $6,420. Elsewhere, stock had been put away one bin to the left of where SAP expected it, so one bin was counted empty and the next bin full. Each mistake looked like a real difference on its own, and they only made sense side by side.

Real differences. That left 37 items with a net loss of $7,410. Some were damaged reels thrown away in December without a scrap posting. Some were small parts that had genuinely gone missing over the year. These are the differences that belong in the books, and the ones worth understanding so they do not happen again.

A year-end count sheet with numbered callouts showing five causes of false differences, from a crossed-out number to a failed goods issue, above the day one totals.

Why posting the whole difference is the expensive shortcut

Posting the $31,240 straight away would have closed the count on Monday morning. It would also have caused problems that took months to unwind.

The book would have been wrong the other way. The 400 couplings were already consumed in production. Posting them as a loss while the failed goods issue was still waiting in COGI would have removed them twice, the moment someone cleared the error list. The same goes for every receipt posted after the count. Writing off a shortage that is really an unposted movement turns a timing difference into a permanent one.

The financial statements would have been wrong too. An inventory loss of $31,240 is a different story from $7,410 of real shrinkage, and it lands in the year's results, in the cost of goods sold, and in whatever measures the plant is judged on. Correcting it after the year is closed means adjustments in the new year that somebody will have to explain.

Planning would have suffered. MRP plans from the stock in SAP. If the system thinks 1,152 brass fittings disappeared, it orders more. If it thinks a pallet of 240 fittings appeared from nowhere, it orders fewer than it should. Wrong counts posted in January become shortages and excess stock in February.

And the causes would have stayed. A goods issue that failed on December 31 will fail again next month for the same reason. A unit of measure that confuses one count team will confuse the next. If differences are posted without a cause, the same items differ again at the next count, and the inventory difference account grows every year without anyone knowing why.

The auditors care about exactly this. They do not just want the count to be accurate. They want to see that differences were investigated, that large write-offs were approved by the right people, and that the explanation for each one holds up. Our post on what auditors look for in an automated SAP audit trail covers why that record matters as much as the number.

A better count day, step by step

Run the same count again, with the investigation built into the process instead of bolted on afterwards.

Step 1. Read the counts as they come in, not on Monday

Count sheets are photographed or scanned as each team finishes an aisle, and scanner exports are picked up as they are saved. The counts are read straight from the sheets, including handwritten numbers and corrections, and matched to the right physical inventory document, material, bin and batch. Where a number has been crossed out and rewritten, both versions are kept, and the one that counts is flagged for a person to confirm. Nobody types 3,412 counts from a cardboard box.

Step 2. Check the usual causes before calling anything a difference

For every item that disagrees with SAP, the usual explanations are checked first. Are there open goods movements for this material, such as failed goods issues in COGI, receipts not yet posted, or open transfer orders? Was it counted in a different unit from the one SAP uses? Does the neighboring bin have the opposite difference? Does the same material and batch appear on two count teams' sheets for the same location? Was the item touched by a movement after the book quantity was frozen?

In the Wisconsin count, this check explains 249 of the 286 differences on Saturday afternoon, while the count teams are still in the building.

Step 3. Send recounts while people are still there

Where the check cannot explain a difference, or where the explanation needs confirming, a recount goes back to a count team the same day. Count bin 02-11-07 again, this time in reels. Check whether the pallet on the zone boundary was counted by team 4 or team 7. Recounting on count day costs minutes. Recounting the following Thursday means reopening the area and stopping movements again.

Step 4. Fix the timing items at their source

Timing differences are not posted as gains and losses. They are fixed where they started. The failed goods issue for production order 1004518 is cleared from COGI, which takes the 400 couplings out of stock properly. The receipt from December 31 is posted with the right date. Once those movements post, the differences disappear without anyone writing anything off.

Step 5. Post only the real differences, with approval

The 37 real differences, $7,410 net, go to the inventory controller with the evidence for each one: the count, any recount, the checks that came back clean, and a photo of the count sheet. Anything above the company's value threshold also goes to the plant controller. Once approved, the differences post through MI07 in the normal way, and the reason for each one is recorded.

The count closes on Monday as planned. The difference that reaches the books is the real one, and every line of it has a reason.

Waterfall chart showing a 31,240 dollar count shortfall reduced to 7,410 dollars of real differences after timing, unit, handwriting, bin and double-count causes are explained.

What an agent checks, and what people still decide

Most of the work in that process is checking, and checking is what an agent does well. It does not get tired at line 212 of a difference list, and it looks at every item the same way.

An agent reads count sheets and scanner exports, including handwriting and crossed-out corrections. It matches each count to the right physical inventory document, material, bin and batch. It checks open goods movements, failed goods issues in COGI, open transfer orders and receipts posted after the count. It compares units of measure, finds the same stock on two teams' sheets, and pairs up neighboring bins with opposite differences. It values each difference and sorts the list by cause, so people see the 37 real problems instead of 286 lines.

What the agent should not do is decide what goes into the books. A person decides whether a crossed-out number should be read as 48 or 84 when the sheet is unclear. A person decides whether to recount or accept. The inventory controller decides which differences are real and approves them. The plant controller approves the large ones, and decides what to do about the causes, whether that is a scrap process nobody follows or a goods issue that keeps failing. The agent prepares each of those decisions with the evidence side by side and records who decided what.

How Artificio handles year-end counts in SAP

Artificio's Inventory Cycle Count Agent is built for exactly this work, whether the count is a weekly cycle count or a year-end count of the whole warehouse. It compares counts with SAP stock and finds out why they differ before any adjustment is posted. Unit of measure, location, batch and open postings are checked first. Real differences post only after the inventory controller approves them, and recounts are requested where the cause is unclear. Outside year end, the same agent decides what to count based on value, movement and past differences, so cycle counts go where errors actually happen.

Counts arrive on paper as often as on scanners, so document capture into SAP and AI OCR read handwritten count sheets and photographed tallies, as well as scanner exports.

Timing differences are the largest category in most counts, and most of them start somewhere else. The COGI item clearance agent clears failed goods movements from production before they turn into phantom shortages. The PO goods receipt agent posts receipts from delivery notes as goods arrive, so stock on the shelf is stock in SAP. Our post on the shop floor confirmation gap explains how late confirmations create the same kind of drift during the year.

Every difference that reaches the books goes to a person through human-in-the-loop review, with the count sheet, the checks and the proposed adjustment side by side. Nothing is written off on its own. Our post on why document AI should never post blind covers the thinking behind that.

Once the count is closed, the real differences are a good place to start looking at what else in the warehouse needs attention. Slow-moving and obsolete items found during the count can go straight to the excess and obsolete inventory agent for the year-end review. And if your GR/IR account is also waiting for attention, our post on the GR/IR account nobody wants to open at year end shows how the same timing problems build up there.

Making next year's count smaller

The best way to make a year-end count easier is to count a little all year. Cycle counting spreads the work across the year, counting each item on a schedule instead of everything on one Saturday. The schedule matters more than the effort. Counting every bin four times a year spends most of the time on items that are almost never wrong.

A better schedule follows value, movement and history. High-value items and fast movers get counted often. Items that differed at the last count go back on the list straight away, because the cause is usually still there. Bins where put-away mistakes happen, and materials that are counted in one unit and stored in another, get extra attention. Slow, cheap and stable items can wait.

The year-end count also becomes a source of information for that schedule. The 37 real differences in Wisconsin, and the 249 explained ones, show exactly where the process leaks. The couplings point to a goods issue that fails at the end of a shift. The brass fittings point to a count sheet that does not say which unit to use. Each one is a reason to count that item, bin or material more often next year.

Companies with reliable cycle counting can sometimes shrink or even replace the full year-end count, if their auditors agree that the counts and controls during the year are good enough. That conversation is much easier with a record of every count, every difference and every explanation.

What to do before this year's count

There are a few weeks left before most year-end counts, and the most useful preparation is unglamorous.

Clear the error lists first. Work through COGI, open transfer orders and unposted receipts in the days before the count. Every failed movement cleared before count day is a phantom difference that never appears.

Print count sheets in the unit people actually count. If a material is stored in boxes, say so on the sheet, with the number of pieces per box. Most unit of measure differences start with a sheet that does not say which unit to count in.

Draw count zones with clear boundaries. Assign every bin to exactly one team, and mark the boundary bins on the floor. Double counts almost always happen at the edges.

Agree the rules for corrections. Ask teams to cross out and initial, never overwrite, and to write the unit next to every number. A correction that is easy to read is a correction that gets keyed right.

Decide who approves what. Set the value limits in SAP for posting differences, name the inventory controller and the plant controller who approve them, and agree how quickly recounts happen on the day.

Read the counts on the day. Whatever tools you use, get counts into SAP and checked while the teams are still in the building. The cheapest recount is the one done an hour after the first count.

A year-end count is one of the few days a year when everyone stops and looks at what is really on the shelves. It deserves a result the business can trust. The difference between posting $31,240 and posting $7,410 is not better counting. It is checking the paperwork before it reaches the books. If you would like to see how this works on your own warehouse and count data, book a short demo and bring last year's difference list.

Lal Singh, SAP AI Automation Expert

CEO & Founder of Artificio

See it in your SAP environment

Request a demo

Bring us a document, a process, or a bottleneck. We'll show how Artificio captures, validates, and posts into SAP — then scale from there.

Request a demo

Security & compliance

Enterprise security across every solution

ISO 27001:2022 certified, SOC 2 Type 2 compliant, GDPR and HIPAA ready. Every agent action is logged, auditable, and runs in isolated environments.

  • ISO 27001:2022
  • SOC 2 Type II
  • GDPR ready
  • HIPAA ready