A stock count is honest only when the counter does not know what the system expects to find. Blind counting hides the expected quantity, variance tolerance filters out noise, and a reason code on every real difference turns a count into evidence you can total. Together they stop the quiet habit of ticking a number that was already on the screen.
Why does a stock count with zero variance prove nothing?
A zero-variance count often means the counter confirmed the number on the screen instead of looking at the shelf. When the expected quantity is visible, the fastest way to finish is to agree with it, and a tired team at 9 pm will take that route every time.
The result is a count that passes, an adjustment of nothing, and a shelf that is still wrong. The error surfaces weeks later as a customer order that cannot be picked, a marketplace cancellation, or a margin number that does not add up. If you already run cycle counting in your warehouse, the schedule is not the weak point. The weak point is what the counter sees on the screen while counting.
Under the Central GST framework, a registered business is expected to keep proper accounts of its stock of goods (see the guidance on cbic-gst.gov.in). A count that only echoes the system is a poor foundation for those records.
What is blind counting?
Blind counting means the counter enters what is physically on the shelf without being shown the system quantity first. The expected number stays hidden from the counter and is visible only to the manager reviewing the finished count.
In Commmerce, an Omnichannel Retail Operating System for Indian retailers, blind counting is switched on per shop in Inventory settings under Stock Take. It is off by default, so no shop's process changes until someone chooses it. When it is on, the running variance is hidden as well, because a variance that updates as you type gives the expected figure away one keystroke at a time. The difference still shows as soon as a line is saved, so the counter sees at once that something is off. Managers reviewing a finished count always see both numbers.
Where blind counting helps most
- Fast movers with many touches: the system number changes all day, so a confirmed number is almost certainly stale.
- High value items: a small error here costs real money.
- Bins that were recently received into: if putaway is delayed, the shelf and the system disagree by design until someone looks.
How do you set a variance tolerance that managers will respect?
Set the tolerance as both a number of units and a percentage, and raise a difference only when it is outside both. This keeps managers focused on gaps that matter and stops them from rubber-stamping a long list of trivial ones.
Two units missing from a pallet of four thousand is noise. Two units missing from a shelf of three is most of the stock. A single flat number cannot tell those apart, which is why a unit allowance and a percentage allowance work better together. In Commmerce both are set per shop in Inventory settings under Stock Take, and both are zero by default, which raises every difference. The count summary then says how many lines need a decision and how many were within tolerance.
Tolerance decides what is put in front of a manager. It does not change the counted figures or the adjustments that follow from them. Start with zero for a month, look at how many lines were raised, and only then loosen it.
Why should every real difference carry a reason?
A reason code tells you whether stock was genuinely lost or whether your records were wrong, and these two have different fixes. Without a reason, every adjustment from a count says only that a count produced it, and nothing can be totalled.
When a difference is raised for review, the manager records what happened to the goods: shrinkage, damage, theft, a count error, a receiving error, a system error or expired. Four of these mean stock was really lost. Three mean the records were wrong. That split is the point. A pile of receiving errors points at the dock, as in transfer shortages caught at the loading bay. A pile of damage points at handling, which connects to stopping damaged stock from being sold as good stock.
The reason is asked only on differences actually being raised. Asking about differences the shop has already said it does not care about is how a field ends up filled in at random. A manager who does not know is not forced to invent a reason.
How to run an honest stock take
An honest stock take hides the expected quantity, filters noise, and records why each real gap happened. Follow these steps in order for one area first, then repeat across the shop.
- Switch on blind counting in Inventory settings under Stock Take for the shop you want to test.
- Set the variance tolerance to zero units and zero percent for the first full count, so every difference is visible.
- Assign each area to a counter who does not handle that area's receiving, so nobody checks their own work.
- Count the shelf and enter the physical quantity before looking at any system figure.
- Review the raised lines as a manager, with both the counted and the expected number side by side.
- Record a reason on every real difference, and leave it unset when nobody knows.
- Approve the count so the stock adjustments are created from it, rather than typed by hand.
- Raise the tolerance a little only after you have seen which lines were pure noise.
What should you do with the reasons once you have them?
Total them by reason, by area and by month, then act on the biggest pile first. A count that shows receiving errors every Monday needs a dock check, not a lecture to counters.
| Reason pattern | What it usually means | First fix |
|---|---|---|
| Repeated receiving errors | Goods booked in without a physical check | Verify quantity against the purchase order at the dock |
| Repeated damage | Handling or storage problem in one area | Look at racking, stacking and who moves the stock |
| Theft or shrinkage in one section | Access or supervision gap | Tighten who can reach that section; see preventing staff theft and shrinkage |
| Frequent count errors | Counters rushing or unclear bin labels | Recount only those lines, improve labels |
| System errors | A process updates stock wrongly | Trace the last few movements of that item |
This is also where chains learn which stores run clean counts. Industry bodies such as the Retailers Association of India regularly highlight inventory accuracy as a core retail discipline, and the sector overview at ibef.org shows why more Indian retailers now sell through several channels from the same stock. The more channels draw on one shelf, the less room there is for a number nobody verified.
What mistakes make blind counts fail?
Blind counts fail when people treat them as a trap instead of a process. Counters who feel punished for every difference will start rounding numbers to avoid questions.
- Setting tolerance too wide on day one: you hide the very pattern you wanted to find.
- Forcing a reason on every line: people invent reasons, and the data becomes useless.
- Counting while receiving goods into the same bin: the count is wrong before it starts. Decide whether work carries on in the area, and count the busiest bins at quiet hours.
- Never looking at the totals: reasons are only useful when someone reads the pile.
- Using it only for annual audits: a short blind count on a few fast movers each week beats one heroic count a year. The method also fits a stock audit that does not shut the store.
Frequently asked questions
Does blind counting slow the count down?
Slightly at first, because counters actually count. It usually pays back, since you recount fewer lines and trust the adjustments that result.
Can a manager still see the expected quantity?
Yes. Only the counter is blind. Managers reviewing a finished count see both the counted and expected numbers.
What is a sensible starting tolerance?
Start at zero so every difference is raised, review one full count, then allow a small number of units and a small percentage based on what you saw.
Should every difference have a reason?
Only the differences actually raised for review. If the manager does not know what happened, the adjustment should record only that a count produced it.
Is this only for warehouses?
No. A store back room or a shelf of fast sellers benefits as much, because the problem is the counter seeing the answer, not the size of the building.
