OĞUZ EROLAI·ADS·APPS

Barcode Inventory Count App: Blind Counting, Location Assignment and Variance Approval

14 min read

Differences found at a year-end stock count come not only from the shelves but from how the count is organised: the same bin is counted by two teams and added together in Excel, a bin missing from the list is never counted, a number of cases is written down as a number of units. The arrangement that prevents this rests on four rules. Each location is assigned to one team and closed once counted. The counter does not see the expected quantity (blind counting). Any variance above a threshold is recounted by a different team. Stock changes only when the variance is approved, through a separate adjustment record. Most of these rules work on paper too; Excel falls short at enforcing them while the count happens. If one or two teams count once a year, you don’t need an application; the cases where you do are below.

Full count or cycle count?

In a full count every location is counted on the same day and stock movements stop; in a cycle count small slices of locations or items are counted each day or week while the warehouse keeps working. Many businesses do a full count at year-end close, when records first move into a new system and when variances are no longer trustworthy. Cycle counting finds differences without waiting a year or closing the warehouse.

In cycle counting, high-value or fast-moving items are counted more often. This classification (ABC) is usually based on annual consumption value; in the example below I split by unit cost for simplicity. Class A items carry 54% of the example’s stock value. Counting them monthly and the rest once or twice a year checks more than half the value every month. The price, roughly and at the example’s rates, is a yearly counting effort of two to three full counts.

The condition for cycle counting is that nothing moves in the location while it is counted. Goods picked but not yet dispatched on paper create a false shortage; the cut-off rule is in step 1 of count day below. Ask your accountant or auditor whether cycle counting can replace the year-end full count where you operate.

Example: 640 locations, 1,480 items, 6 teams

The warehouse and every figure are fictional; I counted the same stock in a Python simulation, and its assumptions are listed under “What I have not verified”.

A distributor of industrial fasteners and consumables runs a warehouse split into 4 zones, 16 aisles and 40 bins per aisle: 640 locations. The bins hold 1,480 items; about 30% arrive in cases. The year-end count is done by 6 teams of two.

When the lists were printed from an old location export, two problems appeared. Both teams counted the aisle-end racks, and because results were summed in Excel, 9 locations doubled. That inflation is about 1.2% of stock value, 38 times the true net count variance. The 14 locations opened after the export were missing from the lists and never counted; two real differences inside them stayed in the records.

In the blind count, 51 rows above the threshold were recounted by another team. On 24 rows the two counts disagreed; the supervisor counted a third time, and on all 24 the first count was wrong. 8 of these errors were case/unit mix-ups. After the recount the net variance was −0.03% of recorded stock value; posting the first count directly would have shown −0.35%. The threshold has a cost: 28 miscounts below it (23 on class C, 5 on class B, ±1-3 units) went into the adjustments.

ProblemWell-built ExcelApplication to be built
Double-counted location (9)Formula flags it after the count; bin is recountedA closed location would not reopen
Missed location (14)Current list and formula find it afterwardsSame current list; unclosed locations would show during the count
Recount (51 rows)Threshold filtered by formula, list printed and handed outRow could drop into another team’s list
Case/unit mix-up (8)Caught at recountCase barcode could convert to units at entry (if the card has a multiplier)
Retyping from paperEvery row typed a second timeCount could be entered on the device

A current list, team assignment and blind counting all work on paper; what an application could add is enforcing these rules during the count and removing the second round of typing.

How does count day run?

Count day runs in seven steps: cut-off, location list, team assignment, blind count, recount, variance approval and adjustment record.

  1. Cut-off. Pending receipts and shipments are posted before count time, then movements stop and each row’s recorded quantity is frozen at that value. How to freeze it in Excel and the order for finding the cause of a difference are in my Excel inventory tracking guide.
  2. Location list. The list is generated from today’s shelf addresses, not an old export. Each bin needs a readable location label. A walk through the warehouse the day before, looking for bins missing from the list, would have found the example’s 14 missed locations.
  3. Team assignment. Locations are split between teams by aisle or zone, and it is written down which team owns the aisle-end racks. In a two-person team one counts, the other scans and records.
  4. Blind count. The counter does not see the recorded quantity. The usual reasoning: a counter who sees the expected figure and finds a close number on the shelf may write the listed figure instead of recounting. I have not measured this effect in the simulation. On paper, a blind count means a list without the recorded quantity column.
  5. Recount threshold. My starting suggestion: every difference on class A, 2 units or more on class B, and on class C 3 units or more and at least 2% of the recorded quantity goes to recount. A different team does the recount, also blind. If the two counts match, the result is accepted; if not, the team lead or warehouse manager counts a third time. It is not a measured value; after the first count, review the differences below it and adjust.
  6. Variance approval. Nobody approves a row counted by their own team. The warehouse manager reviews the difference and the reason; finance separately approves differences above an amount you set.
  7. Adjustment record. The counted quantity is not typed over the stock field. If a row with a recorded quantity of 24 and a count of 22 is approved, a −2 adjustment movement is posted with the count session number and the approver’s name.

What do barcodes and QR codes solve on the floor, and what not?

A barcode speeds up picking the right item and location; it does not fix a miscounted quantity. Four floor decisions follow: the device, case versus unit barcodes, unlabelled items and working without a connection.

Phone camera or handheld scanner?

A web application can open the phone camera only over a secure connection (HTTPS) and only with the user’s permission (MDN: getUserMedia). The browsers’ built-in barcode reading interface (BarcodeDetector) is listed as experimental on MDN and does not work in some of the most widely used browsers (MDN: BarcodeDetector). So camera scanning uses a scanning library or is tested on each target phone. A handheld scanner that behaves like a keyboard types the scanned code into the field on screen and does not depend on browser support.

Light, shelf height, gloves and daily row volume decide the choice. I haven’t measured which device is faster in your warehouse; at the first step I suggest counting one aisle with two devices and comparing. QR codes also work on location labels, as long as item codes and location codes are kept apart.

How are case and unit barcodes kept apart?

Under GS1’s GTIN rules, a unique GTIN is assigned at every packaging level above the base unit, and a change in the number of items in a case requires a new GTIN (GS1: GTIN rules). So the case and the single item carry different barcodes. In the application to be built, each barcode is linked on the product card to one item, one unit and one multiplier; scanning a case barcode would show “1 case = 24 units” and store the count in units. The example’s 8 case/unit errors come from writing the number of cases in the units column. Not every supplier follows GS1 rules; with internal barcodes, you build the mapping yourself.

What about items without a barcode?

Labelling happens before count day, not during it. For items that cannot be labelled, the location barcode is scanned and the item is picked from the list of items recorded in that bin. If an unknown code is scanned, no new product is created; the row goes to review and the counter can attach a photo.

What if the warehouse has no internet?

A web application can open without a connection through a service worker (MDN: offline operation); counts are kept on the device and sent when the connection returns. The browser may not send them on its own (MDN: Background Synchronization). Data stored in the browser can also be evicted under storage pressure unless persistent storage is granted, and Safari, with tracking prevention on, deletes script-created data of a site with no interaction in seven days (MDN: storage quotas).

Rules I suggest for the application to be built: location assignments are downloaded to the device before the count; the number of unsent rows stays visible on screen; a location does not count as “closed” until its rows reach the server; a count for the same location from a second device is rejected and goes to review. Before the count, walk the warehouse with a phone and mark aisles where the connection drops.

Where does a count break in Excel?

Excel holds the counted figure but doesn’t check, while counting, which team counted which bin. It breaks in three places.

  • Two teams count the same location. Everyone fills in their own list, results are combined with SUMIF, and a bin counted twice is added twice.
  • Simultaneous files. A network-folder file not set up for co-authoring usually takes one writer at a time, so each team keeps its own file, and at the end-of-day merge a file can be left out or added twice. Locking and co-authoring are covered in the Excel guide.
  • Blind counting doesn’t hold in a file. Hiding the recorded quantity column isn’t enough; anyone opening the file can unhide it. A blind count needs a separate file or paper list without recorded quantities, and retyping from paper adds another source of error.

If you stay in Excel, add two checks to the combined count table. With the table named Count and columns Location · Item Code · Team · Counted, this flags every row when the same location and item appear more than once:

=COUNTIFS(Count[Location],[@Location],Count[Item Code],[@[Item Code]])>1

To find missed locations, add this column to a Locations table holding every shelf address:

=COUNTIF(Count[Location],[@Location])=0

Enter a count row with quantity 0 for empty bins too; otherwise the second formula flags them as missed. Both formulas find double and missed locations afterwards, not during the count. If a double-counted bin has two different results, only a recount tells which is right.

Where does a web application start, and where is Excel enough?

The number of teams and how often you count make an application necessary, not the number of items; with orderly shelf addresses, a paper list without recorded quantities, a combined Excel table and the two formulas above go a long way.

Run this test in your own warehouse: in the variance list from your last count, separate and count the rows caused by double-counted or missed locations and by case/unit mix-ups. If there are none, stay in Excel. If they make up most of the differences, weigh an application. If you are in between, run one count with paper lists and the two formulas, then look at what remains at recount. Two more situations argue for one:

  • You are moving to cycle counting. Freezing quantities for a new slice, producing the recount list and collecting approvals becomes a daily manual task in Excel.
  • The warehouse has no connection and count results must not get lost on the device.

There is a negative branch too. If your bins are not labelled or the product card has no case multipliers, an application will not solve double counting or case/unit mix-ups; addressing and the product card come first.

ApproachWhenLimit
Blind paper list + Excel1-2 teams, one full count a yearDouble and missed locations found afterwards; retyping from paper
Shared Excel or SheetsSeveral teams, connection availableNo location lock; blind count doesn’t hold in the file
Stock software’s count moduleIf it offers blind counts, recounts and approval, use it firstYour thresholds and roles go only as far as its settings
Custom web applicationMany teams, cycle counts, mixed barcodes, offlineShelf addresses and product cards must be fixed first

Beyond what the table and the sections above cover, the application to be built could include a blind count screen that hides the recorded quantity, role-based variance approval, and export of approved adjustments to a file your stock software can import, after checking that its recorded quantity still matches the frozen one. AI can look at past count differences and suggest which items should enter the cycle count list; the warehouse manager approves the list. Shared functions such as blocking negative stock are on the inventory tracking system page.

It can be built to fit your needs, with the scope defined in writing at the first step. The server can be a VPS in your company’s name; the data is yours and downloadable as Excel or CSV at any time. For small and mid-sized teams there are no per-user licences; the cost is the server plus maintenance. The source code sits in a repository you can access, with run instructions, a dependency list and known limitations.

The first consultation is paid. For a written scope, prepare your number of locations and items, how many teams go out to count and the variance list from your last count, then write through the contact page.

What I could not verify

I read the sources on 31.10.2026.

  • The example is a simulation resting on these assumptions: a real difference on 5% of rows, a 2.5% miscount rate per count, a 1.5% case/unit mix-up rate on cased items, assumed unit costs, independent errors between counts and a third count taken as correct. So the example’s “on all 24 the first count was wrong” is not a measurement; in a real warehouse two teams can make the same mistake.
  • I have not tested the speed or success rate of phone camera scanning on real devices.
  • I have not run the COUNTIFS and COUNTIF formulas in Excel; I tested the logic on the same rows in the simulation.
  • MDN does not state which storage interfaces Safari’s seven-day deletion covers; how long counts stored in the browser last needs testing on the target device.
  • I have not examined whether cycle counting meets statutory year-end inventory requirements in any jurisdiction.

Frequently Asked Questions

What goes to finance or the auditor after the count?

Three records can be prepared. The count list shows which team counted which location and when. The variance list carries, for each row, the recorded and counted quantity, the reason and the approver. The adjustment record lists the movements posted to stock with the count session number. Agree the format of the signed count report and how long to keep it with your accountant or auditor.

What changes if temporary staff join the count?

In the application to be built, a temporary counter could get a user account valid only for that count session, with access ending when the session closes. Every row stores the counter’s name, showing whether recount rows cluster around one person. I suggest pairing each temporary counter with an experienced colleague.

Is consignment or third-party stock in the warehouse counted?

Yes, but reported separately. Location-item rows for stock you don’t own can carry an ownership field. Count variances are split by owner and don’t mix into your own stock value. Settle the accounting treatment with your accountant.

Back to top ↑
WHAT YOU CAN ASK FOR

How do I work?
Work-hour packages

Tell me about your business and what you need on WhatsApp or by phone; we agree on a work block of somewhere between 5 and 40 hours and I start doing what is needed. What you get is not a calendar window — it is net working hours spent directly on your business. I can use 40 hours working three days straight, or deliberately spread it out and finish it over a year — whatever the work calls for. I do not spend my time estimating how long something will take, I spend it doing the work; every hour I spend leaves a lasting result in your business. We keep going for as long as you find it useful.

WORK BLOCKS

Why not monthly?
Is 40 hours a week?

On a fixed monthly fee both sides are watching the wrong thing: you watch a fixed cost, the other side watches how many more contracts like this they can line up. Working by the hour puts us both in front of the same thing — the real value of the work that got done. I know not having an upfront answer to "how much will this cost in total?" is the natural cost of this model. But instead of paying a fixed fee for months to a setup that is not producing results, you only pay for the qualified time spent on your work — and you see exactly where that time went, start to finish. That is why my clients keep coming back.

5hours300€ + VAT60€ per hourChoose this block
10hours500€ + VAT50€ per hourChoose this block
20hours800€ + VAT40€ per hourChoose this block
40hours1,400€ + VAT35€ per hourChoose this block