Inventory control is making sure the stock you already hold is where your records say it is, in the condition your records say, in the quantity your records say.
It is the narrower, more operational half of inventory work. Inventory management decides what to buy and how much to hold. Inventory control looks after what has already arrived.
The distinction sounds academic and has a practical consequence: most businesses that believe they have an inventory management problem actually have an inventory control problem. Sophisticated reorder points calculated from stock figures that are wrong will order the wrong amounts with great precision.
Inventory control vs inventory management
Inventory management is forward-looking and commercial. What to stock, how much, when to reorder, which suppliers, what service level. It deals in forecasts and decisions.
Inventory control is present-tense and operational. Where is it, how much is there, what condition is it in, who moved it. It deals in records and physical reality.
A concrete split:
- Deciding to hold six weeks’ cover of a product — management.
- Knowing you currently have 1,847 units and that 1,847 is correct — control.
- Choosing a supplier with a shorter lead time — management.
- Finding that 40 units were damaged in a rack collision and adjusting the record — control.
Control comes first in any improvement effort. Management decisions are only as good as the data underneath them, and that data is what control produces.
What inventory control involves
Recording every movement. Receipts, put-aways, picks, transfers, adjustments, returns, write-offs. A movement that happens physically without being recorded is where a discrepancy is born, and every one of them will eventually surface.
Location discipline. Every item has a defined home and the system knows it. Without this, “we have 200” is only useful if somebody can find them within a reasonable time.
Counting. Cycle counting continuously, or periodic full counts, to verify records against physical reality.
Investigating discrepancies. Not just adjusting them. An adjustment corrects a number; an investigation finds the process fault that produced it.
Managing condition. Damaged, quarantined, expired and returned stock kept separate from sellable stock, and visible as such.
Controlling access. Who can move stock, who can adjust records, and a record of who did what. Adjustment rights given to everyone produce books that balance and shelves that do not.
Rotation. FIFO or FEFO, enforced by layout rather than instruction.
Reporting. Accuracy, ageing, adjustment volume and reasons, shrinkage.
Why inventory control matters
The short answer is that control sits between two opposing costs. Too little stock and you get stockouts — lost sales, disappointed customers, expedited freight to recover. Too much and your cash flow suffers, because every unit on a shelf is money you cannot spend on anything else. Control is what lets you sit between the two deliberately rather than by accident, and it only works if the numbers are true.
Everything downstream depends on it. Reorder points, safety stock, forecasting, promising delivery dates to customers — all calculate from your stock figures.
Bad data causes visible failures. An order accepted against phantom stock becomes a short shipment, an apology and sometimes a lost customer. The customer experiences a fulfilment failure; the cause was a control failure weeks earlier.
It is where shrinkage shows up. Theft, damage, mis-picks never corrected, administrative error. You cannot address any of it without knowing it is happening.
It protects working capital. Stock you cannot find gets reordered. Operations with poor control routinely carry duplicate inventory they already own.
Audit and compliance. Stock valuation feeds the accounts. In regulated sectors, batch traceability is a legal requirement rather than good practice.
The core techniques
Cycle counting. Count a portion continuously rather than shutting down annually. Count high-value and fast-moving items more often. The benefit is speed of detection — an error found on Tuesday can be traced; the same error found eleven months later is just a number.
Count by location, not by SKU total. A product can reconcile perfectly overall whilst being wrong in two locations that offset each other, and those two locations still cause failed picks.
Barcode scanning at the point of movement. Typed data entry is where most errors originate. Scanning with check digits reduces them by orders of magnitude. The qualifier matters: scanning at a desk twenty minutes later reintroduces everything you were trying to remove.
Perpetual inventory. Records update at the moment of each transaction rather than periodically.
Defined, labelled locations. Aisle, bay, level, position. A logical scheme rather than one only long-serving staff can decode.
Segregated areas for goods-in, quarantine, returns and damaged stock, so unchecked goods cannot be picked.
Adjustment controls. Restrict who can change a stock figure, require a reason code, and review adjustments regularly. A rising adjustment volume is a symptom worth investigating before the count confirms it.
Batch and serial tracking where the goods require it.
ABC-driven effort. Count and control A items tightly, C items lightly. Equal attention across a whole range means insufficient attention on the products that matter.
The systems you can run it on
Inventory control works at every level of sophistication. What changes is how much manual discipline it demands.
Manual counting and memory. No records beyond what somebody can see on the shelf. Workable for a single-room operation with a handful of lines and nothing else. It fails the moment two people are making stock decisions, because neither knows what the other has moved.
Stock cards. A card per product or per location, updated by hand at each movement. Dismissed as antiquated and still in use in plenty of stores and workshops, because it is cheap, it requires no power, and it puts the record physically next to the stock. The weakness is that it records the movement without checking it, so an unrecorded pick leaves no trace at all.
Spreadsheets. The most common starting point for growing businesses. Genuinely capable at small scale, and the ceiling arrives sooner than people expect: one person can edit at a time, there is no audit trail worth the name, and a single mistyped formula propagates silently. Most operations outgrow a spreadsheet somewhere between a few hundred SKUs and the first time two locations need to see the same figure.
Inventory or warehouse software. Records update as transactions happen, multiple users work simultaneously, movements carry an audit trail, and scanning feeds the system directly. The cost is setup discipline — software applied to undefined locations and unreliable processes produces wrong numbers faster than a spreadsheet did.
The progression matters less than the match. A business with 80 SKUs running a well-kept spreadsheet has better control than one with 4,000 SKUs running software nobody updates at the point of movement.
Periodic and perpetual, in practice
The other system distinction is when the record updates.
Periodic means stock figures are established by counting, at intervals — monthly, quarterly, or annually. Between counts, the recorded figure is an estimate derived from purchases and sales. It is simpler and cheaper, and it means that for most of the year you do not actually know what you hold.
Perpetual means every receipt, pick, transfer and adjustment updates the record as it happens. You have a live figure at any moment, which is what makes reorder points, availability promises and automated replenishment possible at all.
Perpetual records still need counting. The record drifts from reality through mis-picks, damage, theft and mis-keyed transactions, and cycle counting is what pulls the two back together. An operation running perpetual records without counting has a live number and no idea whether it is true.
Barcodes and RFID
The technology layer exists for one reason: to remove human transcription from the moment stock moves.
Barcodes encode an identifier that a scanner reads in a fraction of a second, with a check digit that rejects a misread rather than accepting a wrong one. The error rate of scanning is roughly one in several million characters against something closer to one in three hundred for manual keying. They are cheap — a printed label costs almost nothing — and they need line of sight and one scan per item.
RFID uses radio tags that are read without line of sight, often many at once. A pallet can be read as it passes a portal; a shelf can be counted by walking past it. That changes what is practical: full-location counts in minutes rather than hours, and automatic detection of stock in the wrong place.
The trade-offs are straightforward. Tags cost meaningfully more than printed labels, so RFID earns its place on higher-value goods, on reusable containers, or where the speed of counting is itself the benefit. Metal and liquid interfere with reads. And a portal that reads everything within range will happily read the pallet on the next aisle, which is a configuration problem rather than a fault.
Most operations run barcodes throughout and RFID selectively, if at all. The decision is not which technology is better but whether the counting time RFID saves is worth more than the tags cost, and for the majority of general merchandise it is not yet.
Measuring inventory control
Inventory record accuracy (IRA) — the share of counted locations matching exactly.
IRA = (locations matching exactly ÷ locations counted) × 100
Locations, not units, and exact matches, not close ones. Operations that measure by units or allow a tolerance report flattering figures that do not predict pick failures.
Shrinkage — stock lost as a proportion of what you should hold.
Shrinkage = ((book value − counted value) ÷ book value) × 100
Adjustment volume and reasons — how many corrections, and why. The reason codes are where the process faults reveal themselves.
Count coverage — the share of locations counted in a period. High accuracy from counting 5% of locations tells you much less than moderate accuracy from counting 80%.
Dead and ageing stock — reported separately from valuation, because FIFO costing makes old stock look cheaper to write off than it is to replace.
Watch the trend rather than the point. A single month’s accuracy figure is noise; three months moving one way is a signal.
The planning metrics that sit on top
Accuracy measures tell you whether your records are true. A second set of metrics uses those records to decide what to hold and when to buy — and they are only as good as the accuracy underneath them.
Reorder point — the stock level that triggers a purchase. It is the demand you expect during the supplier’s lead time, plus a buffer:
Reorder point = (average daily demand x lead time in days) + safety stock
Calculate the lead time from your last twenty receipts rather than from what the supplier quotes. The gap between the two is a common and entirely avoidable cause of running out.
Economic order quantity (EOQ) — the order size that minimises the combined cost of ordering and holding:
EOQ = square root of ((2 x annual demand x cost per order) / annual holding cost per unit)
Treat it as a sense-check rather than an instruction. The formula assumes steady demand and a fixed ordering cost, neither of which is quite true, and it is useful for showing when your current order size is wildly wrong in one direction.
Inventory turnover — how many times you sell through your average stock in a year:
Turnover = cost of goods sold / average inventory value
Rising turnover with stable service is stock working harder. Rising turnover alongside falling availability is stock being cut too far, which is why it should never be read on its own.
Days of cover — the same figure expressed as time, which most people find easier to act on: average stock divided by average daily usage. Twelve days of cover against a fourteen-day lead time is a problem you can see immediately.
None of these mean anything if your record accuracy is poor. A reorder point calculated from a figure that is wrong 12% of the time fires at the wrong moment 12% of the time.
Investigating a discrepancy properly
The difference between an operation whose accuracy improves and one whose accuracy stays flat is almost entirely what happens after a count finds something wrong.
Recount before adjusting. The second count disagrees with the first more often than people expect. Adjusting on a single count builds errors into the record.
Check the obvious physical causes. Stock in the adjacent location, a pallet put away one bay along, units in a returns cage, a case opened and part-picked without the remainder being recorded. A large proportion of apparent losses are misplacements, and they resolve by looking rather than by adjusting.
Read the transaction history. Most systems record every movement against a location with a timestamp and a user. Working backwards from the last known-correct count usually identifies the transaction where the figure diverged.
Categorise the cause. Mis-pick, receiving error, unrecorded movement, damage, theft, transcription error, system fault. Use a fixed list rather than free text, because free text cannot be counted.
Look for the pattern, not the instance. A single discrepancy is noise. Twenty discrepancies where fourteen are receiving errors on one supplier is a finding you can act on.
Adjust last, and record who authorised it. The adjustment is the closing step, not the response.
This takes longer than adjusting, which is exactly why it gets skipped when the warehouse is busy. The compromise that works is to investigate everything above a value threshold and to sample below it, rather than attempting full investigation of every variance and quietly abandoning the practice within a month.
Who should own inventory control
Accuracy tends to be nobody’s job in the operations that struggle with it, and somebody’s explicit job in the ones that do not.
The warehouse owns the physical reality. Picking correctly, putting away to the right location, recording movements as they happen, flagging damage rather than hiding it.
A designated inventory controller owns the records. Running the cycle count programme, investigating discrepancies, maintaining location data, reporting accuracy. In a smaller operation this is part of somebody’s role rather than a full-time post, but it needs to be named.
Management owns the conditions. Adequate time to count, restricted adjustment rights, and a culture where reporting a mistake is safer than concealing it. This last point does more for accuracy than any technique. Where an error found is treated as a failure to punish, errors stop being found — they simply accumulate until the annual count discovers them all at once, by which time none can be traced.
Two arrangements to avoid. The picker who counts their own locations has an obvious conflict when a discrepancy would reveal their own error. And unrestricted adjustment rights mean a discrepancy can be resolved by whoever finds it, silently, which removes the signal entirely.
Common inventory control failures
Adjusting without investigating. The fastest way to make the numbers agree and the underlying problem persist. Every significant adjustment should carry a reason code, and the reasons should be reviewed.
Counting by SKU rather than location. Hides offsetting errors, which are the ones that cause failed picks.
Deferred data entry. Movements recorded later, from memory or a scribbled note. Every hour of delay adds error.
Unrestricted adjustment rights. If anyone can change a stock figure, discrepancies get quietly corrected rather than investigated, and the signal disappears.
No segregation of returns. Returned stock going straight back onto the shelf without inspection means unsellable goods get picked and shipped again.
Counting only at year end. By then errors are months old, untraceable, and large enough to distort the accounts.
Treating the system as truth. The system holds what it was told. Counting is what tells you whether it was told correctly.
Building a control programme from a standing start
If accuracy is currently unknown or known to be poor, the sequence that works:
Establish a baseline. Count a representative sample of locations and calculate accuracy honestly. It will probably be worse than expected. Record it, because without a starting figure nobody can demonstrate that later effort worked.
Fix the locations before the counting. Every item needs a defined, labelled home. Counting a warehouse where stock lives wherever it fitted produces numbers nobody can act on, and the count itself takes three times as long.
Introduce scanning at receiving first. Goods-in is where errors enter the system and where they are cheapest to prevent. An error caught at receiving costs a minute; the same error caught at picking costs a failed order.
Then scanning at picking, with location confirmation. Scanning the product tells you what was taken. Scanning the location tells you it came from the right place, which is what catches the error before it leaves the building.
Start cycle counting on A items only. Attempting full coverage immediately is how counting programmes get abandoned in the first busy week. A small, sustained programme beats an ambitious one that lapses.
Add reason codes to adjustments and review them monthly from the start, even when volumes are small. The habit is easier to establish early than to introduce later.
Restrict adjustment rights once the process is stable. Doing this too early, before people trust the counting, tends to produce workarounds.
Set PAR levels on consumables. For packaging, labels and anything else consumed rather than sold, a simple minimum-maximum level per location — the PAR level — is quicker to run than a full reorder calculation and stops the stationery cupboard problem of discovering you are out of boxes at four o’clock.
Report accuracy visibly. A figure on a board that the team can see moves behaviour more reliably than a report circulated to management.
Expect this to take a couple of quarters rather than a few weeks. Accuracy improves gradually because it is the accumulation of many small process changes rather than a single intervention, and the gains come mostly from prevention at receiving and picking rather than from finding errors later.
A worked example: a small seller
A business sells kitchenware online. About 400 SKUs, one unit, two staff picking, records kept in the order management system and nobody counting anything.
The symptom. Roughly one order a day ships short or late. Each one takes fifteen minutes of customer service and occasionally a refund. Nobody can say why, because the system says the stock was there.
The baseline. They count 60 locations across their top-selling lines. Thirty-eight match exactly: 63% inventory record accuracy. Of the 22 that do not, 14 hold fewer units than recorded and 8 hold more — which immediately says this is a process problem rather than theft, because theft does not create surplus stock.
What they change, in order.
Every product gets a labelled location — aisle, bay, shelf. That takes a fortnight and it is the largest single piece of work.
Goods-in starts scanning against the purchase order rather than ticking a printed sheet. Two weeks later, receiving errors have largely stopped appearing in the count.
Picking moves to scanning the location and the product. This catches the most common error they had: picking a similar product from the adjacent bay.
They start counting 10 locations a day on their A items, which takes about twenty minutes.
Adjustments get reason codes. Within a month the codes show 40% of corrections are coded “found in wrong location”, which points straight back at the pre-labelling era and at one member of staff putting returns away without scanning.
After one quarter. Accuracy on counted locations is 94%. Short shipments have dropped from about one a day to two or three a month. They have also found roughly $4,000 of stock they had already reordered because it could not be located.
What it cost. Two scanners, a label printer, a fortnight of labelling, and twenty minutes a day. No new software.
The point of the example is the order of operations. Every one of those changes prevents errors at the moment stock moves. None of them involves counting more often, and the counting they do is there to confirm the prevention is working rather than to find the errors.
Where to go deeper
- What is inventory management — the broader discipline this sits within.
- How to manage inventory — the step-by-step process.
- Cycle counting — the main verification technique.
- What is inventory shrinkage — what poor control costs.
- What is a stocktake — the periodic alternative to cycle counting.
- Barcode scanning for inventory — the tool that prevents most errors.
The one-line takeaway
Inventory control is keeping your records and your shelves in agreement — count by location rather than by product total, scan at the point of movement, and investigate discrepancies rather than adjusting them, because everything else you do with inventory calculates from these numbers.
Frequently asked questions
What is inventory control? The practice of ensuring recorded stock matches physical stock in quantity, location and condition. It covers movement recording, location discipline, counting, discrepancy investigation and access control.
What is the difference between inventory control and inventory management? Inventory control concerns the stock you already have — where it is, how much there is, and whether the records are right. Inventory management concerns what to buy, how much to hold and when to reorder. Control produces the data that management decisions depend on, which is why improvement efforts should start there rather than with better ordering rules.
How do I improve inventory accuracy? Scan at the point of movement rather than recording later, count by location through a cycle counting programme, restrict who can adjust stock figures, require reason codes on adjustments, and investigate the patterns in those reasons rather than simply correcting the numbers. Prevention at receiving and picking delivers more than detection afterwards, because an error caught at goods-in costs a minute whilst the same error caught at dispatch costs a failed order and a customer apology.
What is a good inventory record accuracy figure? Operations counting regularly typically report figures in the high nineties measured by exact location match. Below roughly 95% the effects are felt constantly through short picks, emergency counts and orders promised against stock that is not there. What matters more than the absolute number is that it is measured consistently, by location rather than by units, and that the trend is moving the right way over several months rather than bouncing around month to month.
How often should I count stock? Continuously, through cycle counting, rather than annually. Count fast-moving and high-value items monthly or more often, slow-moving low-value items once or twice a year. Full annual counts are increasingly done only where an auditor specifically requires one, because they find errors months after the event when the cause can no longer be traced.
What causes inventory discrepancies? Mis-picks, receiving errors, unrecorded movements, damage not written off, theft, and transcription errors from manual data entry. Misplacement is also common and is frequently mistaken for loss — stock put away one bay along is recorded as missing from where it should be and invisible where it actually is. Reason codes on adjustments are what let you tell which of these you actually have, and the distribution is rarely what people assume before they measure it.
What is phantom inventory? Stock the system believes exists but which is not physically there. It is the most damaging form of inaccuracy because it is invisible until an order is accepted against it and the picker finds an empty location. The usual causes are unrecorded movements, mis-picks never corrected, and damage written off physically but not in the system.
Do I need software for inventory control? Not at very small scale, but the combination of perpetual records and barcode scanning is what removes most manual error, and that needs a system. The point at which spreadsheets stop working is usually when more than one person updates them or you sell through more than one channel.
Warehouse Wise is a free, independent library of guides on inventory management, warehousing and fulfilment, written for the people actually doing the work — new warehouse and inventory staff, small and growing retail and ecommerce businesses, and anyone who holds stock and wants to understand it properly. Every guide explains the term, shows a worked example with real numbers where it helps, and stops. We are independent: we do not sell software or equipment, we take no commission on referrals, and our guides are not ranked by who pays us.