Inventory counts that never reconcile: six sources of book-to-stock mismatches

What most companies do after a stock count is change the books to match the floor, then call it closed. But the discrepancy number itself has no management value. Knowing a part number is off by a few units does not let you make any decision. This post breaks book-to-stock mismatches into six sources on the floor: ship now and do the paperwork later, borrowed materials never returned, returns put in the wrong location, unit conversion, samples and trial builds never deducted, and the hardest one, someone quietly adjusting the books. For each, it covers what trace it leaves and which step you need to add.

Johnny Yang12 min read
Inventory counts that never reconcile: six sources of book-to-stock mismatches

"We count every month, and every month it doesn't match. After the count we change the books to the actual numbers, and then everyone goes home."

That day, on the second floor of the warehouse, the warehouse keeper handed me a count discrepancy report. On the left, the quantity in the system. On the right, the quantity actually counted. The column in the middle was called "Variance". Some positive, some negative, two full printed pages. I asked where these variances came from. He paused and said: "We don't know. That's why it's called variance." The accountant next to him added: "It's like this every month anyway. We adjust it and move on."

I'm Johnny Yang, Brightstream Technology in Taiwan. I've led implementation teams for seven years, handled 252+ projects across 32+ industries. Most warehouses I've walked into are doing the same thing: treating the stock count as a ritual of "making the books match the floor". So let me say it plainly: the inventory discrepancy number itself has close to zero management value. Knowing a part number is off by a few units on the books does not let you make any decision, because you don't know who made it disappear, on which day, or because of which document. Only one thing is useful, and it's called traceability: every discrepancy can be traced back to the transaction that caused it. And on the floor the sources are only a handful: shipping first and doing the paperwork later, borrowed materials never returned, returns put in the wrong location, unit conversions done wrong, samples and trial builds never deducted, and the hardest one to deal with, someone who "helpfully balanced the books for you". Each of these six has a fixed timing and a fixed trace. All of them can be caught. What you should aim for is not zero discrepancy. That's impossible. It's being able to name every single discrepancy.

First, you're right: wanting higher count accuracy is the correct instinct

I'm not going to tell you to stop caring about discrepancies. Owners who care about this have usually been bitten already. Sales sees stock in the system, promises the customer shipment in three days, and the floor turns the place upside down and finds only half. Either the delivery is late or you run a rush job, and the margin is gone on the spot. Or purchasing: the books show material on hand, purchasing doesn't trust it, orders another batch anyway, and slowly the warehouse fills with dead stock nobody dares touch. These pains are real. So wanting higher accuracy is the right direction.

The next step is where it goes wrong. For most companies the next step is counting harder: from annual counts to quarterly, monthly, even hiring someone for cycle counts. Counting more often does let you find discrepancies earlier. It doesn't make them fewer. The reason is simple: a count measures the result. Discrepancies grow in the daily transactions between two counts. Moving the scale into your bedroom and weighing yourself three times a day won't make you lose weight.

So the real question isn't "how do we count more accurately". It's "what happened during this period that nobody recorded".

Six sources, each growing in a gap in the process

Ship first, paperwork later

This is the most common one I've seen, and almost every company thinks it's the exception.

A customer calls in the afternoon, says it's urgent, it has to arrive tomorrow. Sales tells the warehouse, and the goods go out. The order and shipping note get done "tomorrow". Maybe tomorrow they get done. Or the salesperson is traveling, or the shipment got another item added, or whoever does the paperwork remembers the quantity wrong, and it's done incompletely. The goods have left the warehouse. The books still show them.

This discrepancy is easy to recognize: it's almost always negative, actual is less than the books, and it clusters around the few customers who add things last minute and whom sales least dares to refuse. Cross-reference the discrepancy report by customer once, and you usually spot it at a glance.

Fixing it doesn't require a new system. It requires making "ship first" itself an action that leaves a record. Even just a notebook at the warehouse door: what went out, with part number, quantity, destination, and who released it. The next day whoever does the paperwork checks against the notebook, and the discrepancy goes from "no idea where it came from" to "those few from yesterday aren't done yet".

Borrowed materials never returned

The second one happens between the production line and the warehouse, and between departments.

The line is short on material, the foreman comes to the warehouse to borrow some, to be returned when the next delivery comes. Maintenance takes a few spare parts to test, they don't fit, they end up in a toolbox. There's also cross-plant borrowing: this plant is in a rush, borrows from the plant next door, and afterward both sides think the other will handle the books.

The problem with borrowing is that it was "always meant to be returned", so nobody thinks it needs to be recorded at the time. But the return often never happens, or what comes back is a slightly different spec, or it comes back and nobody receives it in. Over time the borrower forgets, the warehouse keeper doesn't remember, and that material hangs on the books forever.

The check is simple: go to the floor and look for a "temporary area". Almost every warehouse has a spot piled with things nobody can say who owns, with handwritten notes stuck on top. The size of that area is roughly the size of your borrowing black hole.

Returns put in the wrong location

Returns are a disaster zone, because they involve three things nobody has defined clearly: has it come back, can it be resold, and where is it.

The most common mistake isn't failing to record it. It's recording it in the wrong location. Good returned goods get put in the inspection area and the books don't move. Or the reverse: something that should be scrapped gets received back into good stock, and the books gain an item that can't actually ship. The first shows up at count time as actual exceeding the books. The second shows on the books but can't be pulled when you need it. They go in opposite directions, and mixed in the same report they cancel each other out. One positive, one negative, exactly offset, the total looks great, but your warehouse has two batches of goods that don't match.

To clean this up, the key is splitting "returned" and "sellable" into two actions and two locations: returned goods go into returns holding first, and only after a decision do they go into good stock or scrap.

Unit conversion

This one is the quietest, and the easiest to dismiss as "normal error".

Purchasing buys by the case, production issues by the piece, sales sells by the set. Two layers of conversion in between. If anyone uses the wrong unit on one document, or the packaging spec changes and the conversion rate doesn't, the books drift. The trouble is the drift is usually a multiple. One mistake is off by a whole case, and then it gets explained as "probably miscounted". Then there's mixing weight and quantity: wire, sheet stock and chemical raw materials are often weighed in and issued by the piece, and the factor in between is often a rule of thumb in some veteran employee's head, adjusted batch by batch. Change the person and you change the formula.

The fix isn't hard. What's hard is that nobody thinks it's a problem: lock down one base unit per part number, convert everything else through conversion rates, keep conversion rates in the master data rather than in someone's head, and log every change. During implementation this is basic data cleanup. It's tedious, but it's the foundation for every number that comes after. If you're evaluating whether to adopt a system, one of the dividing lines in what actually separates an inventory system from ERP is whether it handles multiple units and conversion properly.

Samples and trial builds never deducted

The fifth one shows up in companies that do R&D, prototyping, or send samples to customers.

Engineers take material for trial builds, sales takes finished goods to customers, marketing takes some for photos, the trade show takes a case for display. All of it leaves the warehouse, none of it has a sales order, work order or shipping note, because none of it maps to any revenue. So nobody knows what category to deduct it under, and in the end it isn't deducted. In the projects I've seen, this one is often much bigger in dollar terms than the owner imagines, especially because R&D trial builds are iterative. A batch fails, you build again, the failed batch never gets written off, and at the year-end count it all gets filed as "shrinkage".

To solve it, give each of these activities its own formal issue type: trial build issue, sample shipment, trade show loan. Each one is a document, each one needs approval. The category name doesn't matter. What matters is that it becomes a recorded transaction. Only then can you answer a very practical question: how much do we actually spend on material for prototyping each year.

Someone balanced the books for you

The sixth one is the hardest to talk about, but leaving it out means the list isn't finished.

Month-end close, the books don't match the floor, and someone, to make the numbers pass, goes into the system and changes the stock quantity directly. Not necessarily with bad intent. Usually they just want to get it done: "the actual is this number anyway, just make them match". Or a manager told them to: inventory value can't be off by too much this month, find a way to fix it.

Once someone does this, your books lose all traceability, because every piece of evidence about the source was wiped flat by that "adjustment". Count again next month, the discrepancy is back, adjust again, and it becomes a fixed monthly ritual. The other five sources are still there the whole time, but you'll never find them.

The fix isn't catching people. It's separating permissions from traces. Nobody should be able to change stock quantities directly. Changes should only go through an "inventory gain/loss adjustment" document, with a reason field, a signature, kept in the history. You don't have to ban adjustments. The floor will always have discrepancies whose cause genuinely can't be found. You only need every adjustment to be visible. When the share of "cause unknown" on adjustment documents stays high, that share itself is a management signal.

How to recognize and fix the six sources

SourceWhen it usually happensTrace visible on the floorStep to add
Ship first, paperwork laterRush orders, month-end volume pushGoods in shipping area but no document in systemRecord every shipment release, reconcile next day
Borrowed materials never returnedShortages, cross-department supportUnclaimed material piled in the temporary areaIssue a document at the moment of lending, set a return deadline
Returns put in the wrong locationComplaints, seasonal returnsInspection area mixed with good stockSplit returns and sellable stock into two locations
Unit conversionPackaging change, supplier changeTwo units coexisting for one part numberLock the base unit, keep conversion rates in master data
Samples and trial builds never deductedDevelopment phase, before trade showsOn R&D desks, in sample cabinetsCreate dedicated issue types, require approval
Someone adjusting the booksBefore every month-end closeDiscrepancy report suddenly cleanAdjustment documents only, reason and signature mandatory

Discrepancy rate should not be your KPI

Making the discrepancy rate a KPI directly creates the sixth source. You require the warehouse keeper to keep discrepancies within some range, but there's not much he can do. Most of the real sources aren't in his hands. Sales ships first, production borrows, marketing sends samples. The only variable he controls is the discrepancy report itself. So he learns two things: fudge it at count time, and adjust away the obvious ones before month-end.

In the failed implementations I've seen, quite a few killed the new system exactly this way. Discrepancies spike in the first year after go-live, because everything that used to be smoothed over by hand floats to the surface. The owner sees ugly numbers and starts demanding answers, the pressure lands on the warehouse keeper at the bottom, and the warehouse keeper goes back to the old way of making the numbers look good. Six months later, the inventory in the system is as untrustworthy as the floor, just with an extra interface.

The more reasonable approach is to change the metric from "how much is off" to "how many discrepancies can be traced to a cause". This metric will definitely look bad at first. That's fine. It forces you to fix the gaps in the process instead of forcing people to fix the numbers, and every gap you close moves it up a notch. Progress is visible and can't be faked.

And not every warehouse that doesn't reconcile needs a new system. If you have few SKUs, simple locations, and the same two or three people handling all receiving and shipping, and your discrepancies come mainly from shipping before the paperwork and undeducted samples, what you're missing is discipline, not software. Buying a system at that point just moves a process nobody follows to a more expensive place where nobody keeps following it. On the other hand, if you have multiple locations or plants, multi-unit conversion, outsourced processing going in and out, and returns and exchanges at the same time, people can no longer keep up. You need a tool where every transaction automatically records the time, document number and handler. What you should compare isn't the interface. It's whether you can actually query the transaction history, and whether count adjustments have their own documents and permissions. I break this down in more detail in how to choose an inventory management system.

So what should you do now

No meeting tomorrow. Just do three things.

First, print out the last discrepancy report and split it by part number into two piles, positive and negative. Don't look at the total. Look at the overages and the shortages separately, and they look completely different. Negatives cluster in popular part numbers and a few customers. Positives cluster in returns and items whose packaging changed. This step alone will bring three of the six sources to the surface.

Second, walk the warehouse and find that temporary area. Ask about each item one by one: whose is it, when was it put there, how long is it staying. This takes about an afternoon, but it tells you where the borrowing and returns lines are broken.

Third, check in the system: who can change stock quantities directly, how many times they've done it in the past six months, and whether they wrote a reason. If you can't find out, that's the first thing to fix. Before filling in any process, make "changing the books" leave a trace. Otherwise every investigation after this breaks at the same point.

Once these three are done, the discrepancy report goes from a report nobody wants to read to a list you can assign work from. If you still can't make sense of it, talking for thirty minutes with someone who has done implementations is usually more useful than counting the warehouse three more times yourself.

Found this useful? Pass it on
30 minutes · free

Bring the one thing you are stuck on — a free 30-minute diagnostic

You do not need to know what system you want. Tell Johnny Yang about the one thing that keeps blocking you, and in 30 minutes you will know whether to skip it, build it with coaching, or pay for a custom build. No pitch — if we cannot help, you will hear that straight.

Johnny Yang replies himself · limited spots each month · or just book a free 30-minute diagnostic

Johnny Yang (熊董)
Johnny Yang (熊董)
Founder, Brightstream Technology

Serial founder leading a small team that builds systems into an owner's everyday work.

As a founder
7 yrs
Projects delivered
252+
Industries served
32+
About Johnny